ARM: Fix reset in SPL if SYSRESET is not used
Subsystems (3)
cover ARM: Fix reset in SPL if SYSRESET is not used
Hello,
continuing on the discussion around Claudius' patch for fixing reset in SPL [1] we have taken a closer look at the issue. To quickly summarize the situation:
The original patch was to enable the generic ARM implementation of do_reset if CONFIG_SYSRESET is not enabled in SPL. This would break compilation for some boards which define their own do_reset in board/*/spl.c.
To be more specific, the following 4 boards have this custom do_reset:
- toradex/verdin-imx8mm
- freescale/imx8mn_evk
- freescale/imx8mm_evk
- freescale/imx8mp_evk
I hope we can all agree that do_reset is not at all meant to be implemented in board files. From looking at the related code for imx8m, it feels like this was just a workaround hack to archieve the same thing which Claudius has fixed. So this patch series reverts the addition of do_reset implementations in imx8m board files and instead switches to using the proper fix provided by Claudius.
Additionally, the custom do_reset implementations were passing an address (WDOG1_BASE_ADDR) to reset_cpu() instead of 0. This is the only place in the entire U-Boot tree where this happens. Instead, all other implementations simply ignore the parameter and 0 is passed by callers. It looks a lot like this is a legacy left-over which makes me think that using it for a (hard-coded) watchdog address is not a good idea as it breaks convention with the rest of U-Boot.
[1]: https://patchwork.ozlabs.org/patch/1201959
Claudius Heine (3):
ARM: reset: use do_reset in SPL/TPL if SYSRESET was not enabled for
them
imx: imx8m*: Remove do_reset from board files
imx: imx8m: Don't use the addr parameter of reset_cpu
arch/arm/lib/Makefile | 2 +- arch/arm/mach-imx/imx8m/soc.c | 5 +---- board/freescale/imx8mm_evk/spl.c | 9 --------- board/freescale/imx8mn_evk/spl.c | 9 --------- board/freescale/imx8mp_evk/spl.c | 9 --------- board/toradex/verdin-imx8mm/spl.c | 9 --------- 6 files changed, 2 insertions(+), 41 deletions(-)
Signature · 5 lines
Harald DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-62 Fax: +49-8142-66989-80 Email: hws at denx.de
PATCH 1/3 ARM: reset: use do_reset in SPL/TPL if SYSRESET was not enabled for them
From: Claudius Heine <[email protected]>
In case CONFIG_SYSRESET is set, do_reset from reset.c will not be available anywere, even if SYSRESET is disabled for SPL/TPL.
'do_reset' is called from SPL for instance from the panic handler and PANIC_HANG is not set
--- arch/arm/lib/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/lib/Makefile +1 -1
@@ -57,7 +57,7 @@ obj-y += interrupts_64.o else obj-y += interrupts.o endif -ifndef CONFIG_SYSRESET +ifndef CONFIG_$(SPL_TPL_)SYSRESET obj-y += reset.o endif
Signature · 1 line
2.24.1
PATCH 1/3 ARM: reset: use do_reset in SPL/TPL if SYSRESET was not enabled for them
Quoted 9 lines · Harald Seiler
From: Claudius Heine <[email protected]>
In case
CONFIG_SYSRESETis set,do_resetfromreset.cwill not be available anywere, even if SYSRESET is disabled for SPL/TPL.'
do_reset' is called from SPL for instance from the panic handler andPANIC_HANGis not setSigned-off-by Claudius Heine <[email protected]>
PATCH 1/3 ARM: reset: use do_reset in SPL/TPL if SYSRESET was not enabled for them
In reply to Harald SeilerQuoted 7 lines
From: Claudius Heine <[email protected]>
In caseCONFIG_SYSRESETis set,do_resetfromreset.cwill not be available anywere, even if SYSRESET is disabled for SPL/TPL.
'do_reset' is called from SPL for instance from the panic handler andPANIC_HANGis not setSigned-off-by Claudius Heine <[email protected]>Reviewed-by Marek Vasut <[email protected]>
Applied to u-boot-imx, master, thanks !
Best regards,
Stefano Babic
Signature · 5 lines
===================================================================== DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de =====================================================================
PATCH 2/3 imx: imx8m*: Remove do_reset from board files
In reply to Harald SeilerFrom: Claudius Heine <[email protected]>
Use the do_reset implementation of arch/arm/lib/reset.c in SPL instead. It is very close to what is done here, anyway, and plays more nicely with the rest of U-Boot than adding a custom do_reset implementation into board files.
do_reset from arch/arm/lib/reset.c calls reset_cpu with 0 as the addr parameter while the boards are passing WDOG1_BASE_ADDR. This is ok because the reset_cpu implementation uses WDOG1_BASE_ADDR by default if 0 is passed in.
--- board/freescale/imx8mm_evk/spl.c | 9 --------- board/freescale/imx8mn_evk/spl.c | 9 --------- board/freescale/imx8mp_evk/spl.c | 9 --------- board/toradex/verdin-imx8mm/spl.c | 9 --------- 4 files changed, 36 deletions(-)
board/freescale/imx8mm_evk/spl.c +0 -9
@@ -161,12 +161,3 @@ void board_init_f(ulong dummy)board_init_r (NULL ,0 ); } - -int do_reset (cmd_tbl_t *cmdtp,int flag,int argc,char *const argv[]) -{ -puts ("resetting ...\n" ); - -reset_cpu (WDOG1_BASE_ADDR ); - -return 0 ; -}
board/freescale/imx8mn_evk/spl.c +0 -9
@@ -114,12 +114,3 @@ void board_init_f(ulong dummy)board_init_r (NULL ,0 ); } - -int do_reset (cmd_tbl_t *cmdtp,int flag,int argc,char *const argv[]) -{ -puts ("resetting ...\n" ); - -reset_cpu (WDOG1_BASE_ADDR ); - -return 0 ; -}
board/freescale/imx8mp_evk/spl.c +0 -9
@@ -149,12 +149,3 @@ void board_init_f(ulong dummy)board_init_r (NULL ,0 ); } - -int do_reset (cmd_tbl_t *cmdtp,int flag,int argc,char *const argv[]) -{ -puts ("resetting ...\n" ); - -reset_cpu (WDOG1_BASE_ADDR ); - -return 0 ; -}
board/toradex/verdin-imx8mm/spl.c +0 -9
@@ -169,12 +169,3 @@ void board_init_f(ulong dummy)board_init_r (NULL ,0 ); } - -int do_reset (cmd_tbl_t *cmdtp,int flag,int argc,char *const argv[]) -{ -puts ("resetting ...\n" ); - -reset_cpu (WDOG1_BASE_ADDR ); - -return 0 ; -}
Signature · 1 line
2.24.1
PATCH 2/3 imx: imx8m*: Remove do_reset from board files
Quoted 15 lines · Harald Seiler
From: Claudius Heine <[email protected]>
Use the
do_resetimplementation ofarch/arm/lib/reset.cin SPL instead. It is very close to what is done here, anyway, and plays more nicely with the rest of U-Boot than adding a customdo_resetimplementation into board files.
do_resetfromarch/arm/lib/reset.ccallsreset_cpuwith 0 as the addr parameter while the boards are passingWDOG1_BASE_ADDR. This is ok because thereset_cpuimplementation usesWDOG1_BASE_ADDRby default if 0 is passed in.Co-Authored-by Harald Seiler <[email protected]>Signed-off-by Claudius Heine <[email protected]>Signed-off-by Harald Seiler <[email protected]>
Quoted 6 lines
--- board/freescale/imx8mm_evk/spl.c | 9 --------- board/freescale/imx8mn_evk/spl.c | 9 --------- board/freescale/imx8mp_evk/spl.c | 9 --------- board/toradex/verdin-imx8mm/spl.c | 9 --------- 4 files changed, 36 deletions(-)
PATCH 2/3 imx: imx8m*: Remove do_reset from board files
In reply to Harald SeilerQuoted 9 lines
From: Claudius Heine <[email protected]>
Use thedo_resetimplementation ofarch/arm/lib/reset.cin SPL instead. It is very close to what is done here, anyway, and plays more nicely with the rest of U-Boot than adding a customdo_resetimplementation into board files.do_resetfromarch/arm/lib/reset.ccallsreset_cpuwith 0 as the addr parameter while the boards are passingWDOG1_BASE_ADDR. This is ok because thereset_cpuimplementation usesWDOG1_BASE_ADDRby default if 0 is passed in.
Quoted 4 lines
Co-Authored-by Harald Seiler <[email protected]>Signed-off-by Claudius Heine <[email protected]>Signed-off-by Harald Seiler <[email protected]>Reviewed-by Marek Vasut <[email protected]>
Applied to u-boot-imx, master, thanks !
Best regards,
Stefano Babic
Signature · 5 lines
===================================================================== DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de =====================================================================
PATCH 3/3 imx: imx8m: Don't use the addr parameter of reset_cpu
In reply to Harald SeilerFrom: Claudius Heine <[email protected]>
imx8m has the only implementation of reset_cpu which does not ignore the addr parameter and instead gives it some meaning as the base address of watchdog registers. This breaks convention with the rest of U-Boot where the parameter is ignored and callers are passing in 0.
--- arch/arm/mach-imx/imx8m/soc.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
arch/arm/mach-imx/imx8m/soc.c +1 -4
@@ -385,10 +385,7 @@ int ft_system_setup(void *blob, bd_t *bd)#if defined(CONFIG_SPL_BUILD ) || !defined(CONFIG_SYSRESET )void reset_cpu (ulong addr) { -struct watchdog_regs *wdog = (struct watchdog_regs *)addr; - -if (!addr) - wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ; +struct watchdog_regs *wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ;/* Clear WDA to trigger WDOG_B immediately */ writew ((WCR_WDE |WCR_SRS ), &wdog->wcr);
Signature · 1 line
2.24.1
PATCH 3/3 imx: imx8m: Don't use the addr parameter of reset_cpu
Quoted 21 lines · Harald Seiler
From: Claudius Heine <[email protected]>
imx8m has the only implementation of
reset_cpuwhich does not ignore the addr parameter and instead gives it some meaning as the base address of watchdog registers. This breaks convention with the rest of U-Boot where the parameter is ignored and callers are passing in 0.Fixes d2041725e84b ("imx8m: restrict reset_cpu")Co-Authored-by Harald Seiler <[email protected]>Signed-off-by Claudius Heine <[email protected]>Signed-off-by Harald Seiler <[email protected]>--- arch/arm/mach-imx/imx8m/soc.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)arch/arm/mach-imx/imx8m/soc.c +0
-0@@ -385,10 +385,7 @@ int ft_system_setup(void *blob, bd_t *bd)#if defined(CONFIG_SPL_BUILD ) || !defined(CONFIG_SYSRESET )
arch/arm/mach-imx/imx8m/soc.c int ft_system_setup(void *blob, bd_t *bd) · line 391
Quoted 10 lines
arch/arm/mach-imx/imx8m/soc.c +1
-4void reset_cpu (ulong addr) { -struct watchdog_regs *wdog = (struct watchdog_regs *)addr; - -if (!addr) - wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ; +struct watchdog_regs *wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ;/* Clear WDA to trigger WDOG_B immediately */ writew ((WCR_WDE |WCR_SRS ), &wdog->wcr);
Can't we remove the param altogether ?
Otherwise
PATCH 3/3 imx: imx8m: Don't use the addr parameter of reset_cpu
Hello Marek,
Quoted 30 lines · Marek Vasut
Quoted 29 lines · Harald Seiler
From: Claudius Heine <[email protected]>
imx8m has the only implementation of
reset_cpuwhich does not ignore the addr parameter and instead gives it some meaning as the base address of watchdog registers. This breaks convention with the rest of U-Boot where the parameter is ignored and callers are passing in 0.Fixes d2041725e84b ("imx8m: restrict reset_cpu")Co-Authored-by Harald Seiler <[email protected]>Signed-off-by Claudius Heine <[email protected]>Signed-off-by Harald Seiler <[email protected]>--- arch/arm/mach-imx/imx8m/soc.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)arch/arm/mach-imx/imx8m/soc.c +1
-4@@ -385,10 +385,7 @@ int ft_system_setup(void *blob, bd_t *bd)#if defined(CONFIG_SPL_BUILD ) || !defined(CONFIG_SYSRESET )void reset_cpu (ulong addr) { -struct watchdog_regs *wdog = (struct watchdog_regs *)addr; - -if (!addr) - wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ; +struct watchdog_regs *wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR ;
Quoted 4 lines
arch/arm/mach-imx/imx8m/soc.c int ft_system_setup(void *blob, bd_t *bd) · line 391
Quoted 2 lines
arch/arm/mach-imx/imx8m/soc.c +0
-0/* Clear WDA to trigger WDOG_B immediately */ writew ((WCR_WDE |WCR_SRS ), &wdog->wcr);Can't we remove the param altogether ?
Yes, that is what I would suggest as well. Although I'd do that as a separate follow up because it is not ARM nor i.MX specific.
Quoted 2 lines
Otherwise
Reviewed-by Marek Vasut <[email protected]>
Signature · 1 line
Harald
PATCH 3/3 imx: imx8m: Don't use the addr parameter of reset_cpu
In reply to Harald SeilerQuoted 10 lines
From: Claudius Heine <[email protected]>
imx8m has the only implementation ofreset_cpuwhich does not ignore the addr parameter and instead gives it some meaning as the base address of watchdog registers. This breaks convention with the rest of U-Boot where the parameter is ignored and callers are passing in 0.Fixes d2041725e84b ("imx8m: restrict reset_cpu")Co-Authored-by Harald Seiler <[email protected]>Signed-off-by Claudius Heine <[email protected]>Signed-off-by Harald Seiler <[email protected]>Reviewed-by Marek Vasut <[email protected]>
Applied to u-boot-imx, master, thanks !
Best regards,
Stefano Babic
Signature · 5 lines
===================================================================== DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de =====================================================================
cover ARM: Fix reset in SPL if SYSRESET is not used
In reply to Harald SeilerHello,
Quoted 41 lines · Harald Seiler
Hello,
continuing on the discussion around Claudius' patch for fixing reset in SPL [1] we have taken a closer look at the issue. To quickly summarize the situation:
The original patch was to enable the generic ARM implementation of
do_resetifCONFIG_SYSRESETis not enabled in SPL. This would break compilation for some boards which define their owndo_resetinboard/*/spl.c.To be more specific, the following 4 boards have this custom
do_reset:- toradex/verdin-imx8mm
- freescale/imx8mn_evk
- freescale/imx8mm_evk
- freescale/imx8mp_evkI hope we can all agree that
do_resetis not at all meant to be implemented in board files. From looking at the related code for imx8m, it feels like this was just a workaround hack to archieve the same thing which Claudius has fixed. So this patch series reverts the addition ofdo_resetimplementations in imx8m board files and instead switches to using the proper fix provided by Claudius.Additionally, the custom
do_resetimplementations were passing an address (WDOG1_BASE_ADDR) toreset_cpu()instead of 0. This is the only place in the entire U-Boot tree where this happens. Instead, all other implementations simply ignore the parameter and 0 is passed by callers. It looks a lot like this is a legacy left-over which makes me think that using it for a (hard-coded) watchdog address is not a good idea as it breaks convention with the rest of U-Boot.[1]: https://patchwork.ozlabs.org/patch/1201959
Claudius Heine (3):
ARM: reset: usedo_resetin SPL/TPL if SYSRESET was not enabled for
them
imx: imx8m*: Removedo_resetfrom board files
imx: imx8m: Don't use the addr parameter ofreset_cpu
Quoted 8 lines
arch/arm/lib/Makefile | 2 +- arch/arm/mach-imx/imx8m/soc.c | 5 +---- board/freescale/imx8mm_evk/spl.c | 9 --------- board/freescale/imx8mn_evk/spl.c | 9 --------- board/freescale/imx8mp_evk/spl.c | 9 --------- board/toradex/verdin-imx8mm/spl.c | 9 --------- 6 files changed, 2 insertions(+), 41 deletions(-)
Quick question, what is the situation with this series? Is there anything which needs to be addressed?
Without it, CONFIG_SPL_USB_SDP_SUPPORT is broken on the imx6q hardware I am working with and I guess the same issue might exist on many other boards as well (The USB stack needs a do_reset() implementation in SPL).
Regards,
Signature · 1 line
Harald
cover ARM: Fix reset in SPL if SYSRESET is not used
Quoted 53 lines · Harald Seiler
Hello,
Quoted 41 lines · Harald Seiler
Hello,
continuing on the discussion around Claudius' patch for fixing reset in SPL [1] we have taken a closer look at the issue. To quickly summarize the situation:
The original patch was to enable the generic ARM implementation of
do_resetifCONFIG_SYSRESETis not enabled in SPL. This would break compilation for some boards which define their owndo_resetinboard/*/spl.c.To be more specific, the following 4 boards have this custom
do_reset:- toradex/verdin-imx8mm
- freescale/imx8mn_evk
- freescale/imx8mm_evk
- freescale/imx8mp_evkI hope we can all agree that
do_resetis not at all meant to be implemented in board files. From looking at the related code for imx8m, it feels like this was just a workaround hack to archieve the same thing which Claudius has fixed. So this patch series reverts the addition ofdo_resetimplementations in imx8m board files and instead switches to using the proper fix provided by Claudius.Additionally, the custom
do_resetimplementations were passing an address (WDOG1_BASE_ADDR) toreset_cpu()instead of 0. This is the only place in the entire U-Boot tree where this happens. Instead, all other implementations simply ignore the parameter and 0 is passed by callers. It looks a lot like this is a legacy left-over which makes me think that using it for a (hard-coded) watchdog address is not a good idea as it breaks convention with the rest of U-Boot.[1]: https://patchwork.ozlabs.org/patch/1201959
Claudius Heine (3):
ARM: reset: usedo_resetin SPL/TPL if SYSRESET was not enabled for
them
imx: imx8m*: Removedo_resetfrom board files
imx: imx8m: Don't use the addr parameter ofreset_cpuQuoted 8 lines
arch/arm/lib/Makefile | 2 +- arch/arm/mach-imx/imx8m/soc.c | 5 +---- board/freescale/imx8mm_evk/spl.c | 9 --------- board/freescale/imx8mn_evk/spl.c | 9 --------- board/freescale/imx8mp_evk/spl.c | 9 --------- board/toradex/verdin-imx8mm/spl.c | 9 --------- 6 files changed, 2 insertions(+), 41 deletions(-)
Quoted 2 lines
Quick question, what is the situation with this series? Is there anything which needs to be addressed?
Patches are fine IMHO - I will merge them soon.
Stefano
Quoted 7 lines
Without it,
CONFIG_SPL_USB_SDP_SUPPORTis broken on the imx6q hardware I am working with and I guess the same issue might exist on many other boards as well (The USB stack needs ado_reset()implementation in SPL).Regards,
Signature · 5 lines
===================================================================== DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: +49-8142-66989-53 Fax: +49-8142-66989-80 Email: sbabic at denx.de =====================================================================