[U-Boot] Beaglebone Black broken since commit fd61d39970b9901217efc7536d9f3a61b4e1752a
Hi,
Beaglebone black (am335x_evm_defconfig) is broken (with MMC boot and u-boot.img on ext4 partition). I bisected it to commit fd61d39970b9901217efc7536d9f3a61b4e1752a spl: mmc: add break statements in spl_mmc_load_image() from Nikita Kiryanov (in Cc).
Working image gives: ******************************************************************************** U-Boot SPL 2016.01-rc1-00036-g483ab3d (Feb 16 2016 - 19:04:10) bad magic bad magic spl_register_fat_device: fat register err - -1 spl_register_fat_device: fat register err - -1 spl_load_image_fat: error reading image u-boot.img, err - -1 spl: ext4fs_open failed spl: ext4fs_open failed spl_load_image_ext: error reading image uImage, err - -1
U-Boot 2016.01-rc1-00033-g9884f44 (Feb 16 2016 - 18:55:52 +0100) ********************************************************************************
Since this commit, we get: ******************************************************************************** U-Boot SPL 2016.01-rc1-00037-gfd61d39 (Feb 16 2016 - 19:01:54) bad magic bad magic ### ERROR ### Please RESET the board ### ********************************************************************************
Guillaume
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
--- common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif - break; + /* Fall through */ case MMCSD_MODE_FS: debug("spl: mmc boot mode: fs\n");
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */
This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
-- Tom
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior. Then, if someone (you, me or someone else) wants to improve this code, the way you suggested, it would be very nice. But it will need a lot more work, tests and reviews.
Guillaume
-- Tom
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Then, if someone (you, me or someone else) wants to improve this code, the way you suggested, it would be very nice. But it will need a lot more work, tests and reviews.
Guillaume
-- Tom
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean. I think, "Fall through" code comment explains it (as you done with MMCSD_MODE_EMMCBOOT). I (and openSUSE ARM) do not pretend to be the only user(s) of this feature, so I won't add my name there. If you mean commit message, I think the current one is enough.
Guillaume
Then, if someone (you, me or someone else) wants to improve this code, the way you suggested, it would be very nice. But it will need a lot more work, tests and reviews.
Guillaume
-- Tom
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..2eef0f2 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* Fall through */This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black, * depend on this funcitonality. */
I think, "Fall through" code comment explains it (as you done with MMCSD_MODE_EMMCBOOT).
That was meant to communicate a very different type of coupling between the two cases though..
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
>Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: > spl: mmc: add break statements in spl_mmc_load_image() >RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >board hangs. This patch allows to try MMCSD_MODE_FS then, if available. > >It has been tested on a beaglebone black to boot on an EXT partition. > >Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >Cc: Tom Rini trini@konsulko.com >Cc: Nikita Kiryanov nikita@compulab.co.il >Cc: Igor Grinberg grinberg@compulab.co.il >Cc: Paul Kocialkowski contact@paulk.fr >Cc: Pantelis Antoniou panto@antoniou-consulting.com >Cc: Simon Glass sjg@chromium.org >Cc: Matwey V. Kornilov matwey.kornilov@gmail.com > >--- > common/spl/spl_mmc.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > >diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >index c3931c6..2eef0f2 100644 >--- a/common/spl/spl_mmc.c >+++ b/common/spl/spl_mmc.c >@@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) > if (!err) > return err; > #endif >- break; >+ /* Fall through */ > case MMCSD_MODE_FS: > debug("spl: mmc boot mode: fs\n"); This also essentially reverts fd61d399. So Nikita, was there a specific use case that was broken before, or was the code just unclear in intentions here? Thanks!
There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific. It's saying that instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
On Thu, Feb 18, 2016 at 09:36:01AM -0500, Tom Rini wrote:
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
Hi Tom, Guillaume,
On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote: >On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote: > >>Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: >> spl: mmc: add break statements in spl_mmc_load_image() >>RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >>board hangs. This patch allows to try MMCSD_MODE_FS then, if available. >> >>It has been tested on a beaglebone black to boot on an EXT partition. >> >>Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >>Cc: Tom Rini trini@konsulko.com >>Cc: Nikita Kiryanov nikita@compulab.co.il >>Cc: Igor Grinberg grinberg@compulab.co.il >>Cc: Paul Kocialkowski contact@paulk.fr >>Cc: Pantelis Antoniou panto@antoniou-consulting.com >>Cc: Simon Glass sjg@chromium.org >>Cc: Matwey V. Kornilov matwey.kornilov@gmail.com >> >>--- >> common/spl/spl_mmc.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >>diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >>index c3931c6..2eef0f2 100644 >>--- a/common/spl/spl_mmc.c >>+++ b/common/spl/spl_mmc.c >>@@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) >> if (!err) >> return err; >> #endif >>- break; >>+ /* Fall through */ >> case MMCSD_MODE_FS: >> debug("spl: mmc boot mode: fs\n"); >This also essentially reverts fd61d399. So Nikita, was there a specific >use case that was broken before, or was the code just unclear in >intentions here? Thanks! There was no broken use case that I'm aware of. The change was made as part of a code improvement series and was meant to address what I consider to be bad and problematic design. Instead of reverting it though, how about implementing something similar to what I did in the main common/spl/spl.c:board_init_r()? You would have a weak function that will default to the original spl_boot_mode() if not overridden, and allow the user to define a sequence of boot modes otherwise.
The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific.
The comment I proposed does not suggest it's board specific, just that this specific use case is used by someone.
instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
I don't think that's clear enough that this is the purpose of the missing break statement. It's a little too implicit. What I'm suggesting is that we make it a bit more explicit, barring a rewrite.
-- Tom
Le 18/02/2016 17:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 09:36:01AM -0500, Tom Rini wrote:
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote:
Hi Tom, Nikita ,
Le 18/02/2016 10:19, Nikita Kiryanov a écrit : > Hi Tom, Guillaume, > > On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote: >> On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote: >> >>> Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: >>> spl: mmc: add break statements in spl_mmc_load_image() >>> RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >>> board hangs. This patch allows to try MMCSD_MODE_FS then, if available. >>> >>> It has been tested on a beaglebone black to boot on an EXT partition. >>> >>> Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >>> Cc: Tom Rini trini@konsulko.com >>> Cc: Nikita Kiryanov nikita@compulab.co.il >>> Cc: Igor Grinberg grinberg@compulab.co.il >>> Cc: Paul Kocialkowski contact@paulk.fr >>> Cc: Pantelis Antoniou panto@antoniou-consulting.com >>> Cc: Simon Glass sjg@chromium.org >>> Cc: Matwey V. Kornilov matwey.kornilov@gmail.com >>> >>> --- >>> common/spl/spl_mmc.c | 2 +- >>> 1 file changed, 1 insertion(+), 1 deletion(-) >>> >>> diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >>> index c3931c6..2eef0f2 100644 >>> --- a/common/spl/spl_mmc.c >>> +++ b/common/spl/spl_mmc.c >>> @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) >>> if (!err) >>> return err; >>> #endif >>> - break; >>> + /* Fall through */ >>> case MMCSD_MODE_FS: >>> debug("spl: mmc boot mode: fs\n"); >> This also essentially reverts fd61d399. So Nikita, was there a specific >> use case that was broken before, or was the code just unclear in >> intentions here? Thanks! > There was no broken use case that I'm aware of. The change was made as > part of a code improvement series and was meant to address what I > consider to be bad and problematic design. Instead of reverting it > though, how about implementing something similar to what I did in the > main common/spl/spl.c:board_init_r()? You would have a weak function > that will default to the original spl_boot_mode() if not overridden, > and allow the user to define a sequence of boot modes otherwise. The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific.
The comment I proposed does not suggest it's board specific, just that this specific use case is used by someone.
instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
I don't think that's clear enough that this is the purpose of the missing break statement. It's a little too implicit. What I'm suggesting is that we make it a bit more explicit, barring a rewrite.
So, maybe just: /* If raw mode fails, try fs mode. */ ?
Guillaume
-- Tom
On Thu, Feb 18, 2016 at 05:11:46PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 17:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 09:36:01AM -0500, Tom Rini wrote:
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote: >Hi Tom, Nikita , > >Le 18/02/2016 10:19, Nikita Kiryanov a écrit : >>Hi Tom, Guillaume, >> >>On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote: >>>On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote: >>> >>>>Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: >>>> spl: mmc: add break statements in spl_mmc_load_image() >>>>RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >>>>board hangs. This patch allows to try MMCSD_MODE_FS then, if available. >>>> >>>>It has been tested on a beaglebone black to boot on an EXT partition. >>>> >>>>Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >>>>Cc: Tom Rini trini@konsulko.com >>>>Cc: Nikita Kiryanov nikita@compulab.co.il >>>>Cc: Igor Grinberg grinberg@compulab.co.il >>>>Cc: Paul Kocialkowski contact@paulk.fr >>>>Cc: Pantelis Antoniou panto@antoniou-consulting.com >>>>Cc: Simon Glass sjg@chromium.org >>>>Cc: Matwey V. Kornilov matwey.kornilov@gmail.com >>>> >>>>--- >>>> common/spl/spl_mmc.c | 2 +- >>>> 1 file changed, 1 insertion(+), 1 deletion(-) >>>> >>>>diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >>>>index c3931c6..2eef0f2 100644 >>>>--- a/common/spl/spl_mmc.c >>>>+++ b/common/spl/spl_mmc.c >>>>@@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) >>>> if (!err) >>>> return err; >>>> #endif >>>>- break; >>>>+ /* Fall through */ >>>> case MMCSD_MODE_FS: >>>> debug("spl: mmc boot mode: fs\n"); >>>This also essentially reverts fd61d399. So Nikita, was there a specific >>>use case that was broken before, or was the code just unclear in >>>intentions here? Thanks! >>There was no broken use case that I'm aware of. The change was made as >>part of a code improvement series and was meant to address what I >>consider to be bad and problematic design. Instead of reverting it >>though, how about implementing something similar to what I did in the >>main common/spl/spl.c:board_init_r()? You would have a weak function >>that will default to the original spl_boot_mode() if not overridden, >>and allow the user to define a sequence of boot modes otherwise. >The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior. Could you add a comment indicating that this is wanted behavior that has a user, and who the user is?
Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific.
The comment I proposed does not suggest it's board specific, just that this specific use case is used by someone.
instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
I don't think that's clear enough that this is the purpose of the missing break statement. It's a little too implicit. What I'm suggesting is that we make it a bit more explicit, barring a rewrite.
So, maybe just: /* If raw mode fails, try fs mode. */ ?
That'll work too.
Guillaume
-- Tom
Le 18/02/2016 17:38, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 05:11:46PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 17:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 09:36:01AM -0500, Tom Rini wrote:
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 14:07, Nikita Kiryanov a écrit : > On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote: >> Hi Tom, Nikita , >> >> Le 18/02/2016 10:19, Nikita Kiryanov a écrit : >>> Hi Tom, Guillaume, >>> >>> On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote: >>>> On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote: >>>> >>>>> Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: >>>>> spl: mmc: add break statements in spl_mmc_load_image() >>>>> RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >>>>> board hangs. This patch allows to try MMCSD_MODE_FS then, if available. >>>>> >>>>> It has been tested on a beaglebone black to boot on an EXT partition. >>>>> >>>>> Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >>>>> Cc: Tom Rini trini@konsulko.com >>>>> Cc: Nikita Kiryanov nikita@compulab.co.il >>>>> Cc: Igor Grinberg grinberg@compulab.co.il >>>>> Cc: Paul Kocialkowski contact@paulk.fr >>>>> Cc: Pantelis Antoniou panto@antoniou-consulting.com >>>>> Cc: Simon Glass sjg@chromium.org >>>>> Cc: Matwey V. Kornilov matwey.kornilov@gmail.com >>>>> >>>>> --- >>>>> common/spl/spl_mmc.c | 2 +- >>>>> 1 file changed, 1 insertion(+), 1 deletion(-) >>>>> >>>>> diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >>>>> index c3931c6..2eef0f2 100644 >>>>> --- a/common/spl/spl_mmc.c >>>>> +++ b/common/spl/spl_mmc.c >>>>> @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) >>>>> if (!err) >>>>> return err; >>>>> #endif >>>>> - break; >>>>> + /* Fall through */ >>>>> case MMCSD_MODE_FS: >>>>> debug("spl: mmc boot mode: fs\n"); >>>> This also essentially reverts fd61d399. So Nikita, was there a specific >>>> use case that was broken before, or was the code just unclear in >>>> intentions here? Thanks! >>> There was no broken use case that I'm aware of. The change was made as >>> part of a code improvement series and was meant to address what I >>> consider to be bad and problematic design. Instead of reverting it >>> though, how about implementing something similar to what I did in the >>> main common/spl/spl.c:board_init_r()? You would have a weak function >>> that will default to the original spl_boot_mode() if not overridden, >>> and allow the user to define a sequence of boot modes otherwise. >> The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior. > Could you add a comment indicating that this is wanted behavior that > has a user, and who the user is? Not sure what you mean.
I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific.
The comment I proposed does not suggest it's board specific, just that this specific use case is used by someone.
instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
I don't think that's clear enough that this is the purpose of the missing break statement. It's a little too implicit. What I'm suggesting is that we make it a bit more explicit, barring a rewrite.
So, maybe just: /* If raw mode fails, try fs mode. */ ?
That'll work too.
If Tom is ok, I will send a V2.
Guillaume
On Thu, Feb 18, 2016 at 05:42:29PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 17:38, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 05:11:46PM +0100, Guillaume Gardet wrote:
Le 18/02/2016 17:07, Nikita Kiryanov a écrit :
On Thu, Feb 18, 2016 at 09:36:01AM -0500, Tom Rini wrote:
On Thu, Feb 18, 2016 at 04:25:29PM +0200, Nikita Kiryanov wrote:
On Thu, Feb 18, 2016 at 02:31:08PM +0100, Guillaume Gardet wrote: >Le 18/02/2016 14:07, Nikita Kiryanov a écrit : >>On Thu, Feb 18, 2016 at 11:06:32AM +0100, Guillaume Gardet wrote: >>>Hi Tom, Nikita , >>> >>>Le 18/02/2016 10:19, Nikita Kiryanov a écrit : >>>>Hi Tom, Guillaume, >>>> >>>>On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote: >>>>>On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote: >>>>> >>>>>>Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: >>>>>> spl: mmc: add break statements in spl_mmc_load_image() >>>>>>RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the >>>>>>board hangs. This patch allows to try MMCSD_MODE_FS then, if available. >>>>>> >>>>>>It has been tested on a beaglebone black to boot on an EXT partition. >>>>>> >>>>>>Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr >>>>>>Cc: Tom Rini trini@konsulko.com >>>>>>Cc: Nikita Kiryanov nikita@compulab.co.il >>>>>>Cc: Igor Grinberg grinberg@compulab.co.il >>>>>>Cc: Paul Kocialkowski contact@paulk.fr >>>>>>Cc: Pantelis Antoniou panto@antoniou-consulting.com >>>>>>Cc: Simon Glass sjg@chromium.org >>>>>>Cc: Matwey V. Kornilov matwey.kornilov@gmail.com >>>>>> >>>>>>--- >>>>>> common/spl/spl_mmc.c | 2 +- >>>>>> 1 file changed, 1 insertion(+), 1 deletion(-) >>>>>> >>>>>>diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c >>>>>>index c3931c6..2eef0f2 100644 >>>>>>--- a/common/spl/spl_mmc.c >>>>>>+++ b/common/spl/spl_mmc.c >>>>>>@@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) >>>>>> if (!err) >>>>>> return err; >>>>>> #endif >>>>>>- break; >>>>>>+ /* Fall through */ >>>>>> case MMCSD_MODE_FS: >>>>>> debug("spl: mmc boot mode: fs\n"); >>>>>This also essentially reverts fd61d399. So Nikita, was there a specific >>>>>use case that was broken before, or was the code just unclear in >>>>>intentions here? Thanks! >>>>There was no broken use case that I'm aware of. The change was made as >>>>part of a code improvement series and was meant to address what I >>>>consider to be bad and problematic design. Instead of reverting it >>>>though, how about implementing something similar to what I did in the >>>>main common/spl/spl.c:board_init_r()? You would have a weak function >>>>that will default to the original spl_boot_mode() if not overridden, >>>>and allow the user to define a sequence of boot modes otherwise. >>>The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior. >>Could you add a comment indicating that this is wanted behavior that >>has a user, and who the user is? >Not sure what you mean. I mean something like: /* If raw mode fails, try fs mode. Some boards, such as beaglebone black,
- depend on this funcitonality.
*/
But it's not board specific, it's use-case specific.
The comment I proposed does not suggest it's board specific, just that this specific use case is used by someone.
instead of shoving both SPL and U-Boot into the correct magic raw location, just shove SPL there and let U-Boot itself be in the /boot partition. This is just as applicable on say imx6 as it is on TI parts.
I don't think that's clear enough that this is the purpose of the missing break statement. It's a little too implicit. What I'm suggesting is that we make it a bit more explicit, barring a rewrite.
So, maybe just: /* If raw mode fails, try fs mode. */ ?
That'll work too.
If Tom is ok, I will send a V2.
OK with me, thanks!
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
--- Changes in V2: - Replace "/* Fall through */" comment by: "/* If RAW mode fails, try FS mode. */"
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..c27a250 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif - break; + /* If RAW mode fails, try FS mode. */ case MMCSD_MODE_FS: debug("spl: mmc boot mode: fs\n");
On Thu, Feb 18, 2016 at 06:17:36PM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then.
It has been tested on a beaglebone black to boot on an EXT partition.
Acked-by: Nikita Kiryanov nikita@compulab.co.il
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com
Changes in V2:
- Replace "/* Fall through */" comment by:
"/* If RAW mode fails, try FS mode. */"
common/spl/spl_mmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c index c3931c6..c27a250 100644 --- a/common/spl/spl_mmc.c +++ b/common/spl/spl_mmc.c @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device) if (!err) return err; #endif
break;
/* If RAW mode fails, try FS mode. */-- 1.8.4.5
On Thu, Feb 18, 2016 at 06:17:36PM +0100, Guillaume GARDET wrote:
Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a: spl: mmc: add break statements in spl_mmc_load_image() RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the board hangs. This patch allows to try MMCSD_MODE_FS then.
It has been tested on a beaglebone black to boot on an EXT partition.
Signed-off-by: Guillaume GARDET guillaume.gardet@free.fr Cc: Tom Rini trini@konsulko.com Cc: Nikita Kiryanov nikita@compulab.co.il Cc: Igor Grinberg grinberg@compulab.co.il Cc: Paul Kocialkowski contact@paulk.fr Cc: Pantelis Antoniou panto@antoniou-consulting.com Cc: Simon Glass sjg@chromium.org Cc: Matwey V. Kornilov matwey.kornilov@gmail.com Acked-by: Nikita Kiryanov nikita@compulab.co.il
Applied to u-boot/master, thanks!
participants (4)
-
Guillaume GARDET -
Guillaume Gardet
-
Nikita Kiryanov -
Tom Rini