powerpc: fix regression in arch_initr_trap()
Subsystems (5)
PATCH 1/1 powerpc: fix regression in arch_initr_trap()
The assembly output of the arch_initr_trap() function differed by a single byte after common.h was removed from traps.c:
fff49a18 <arch_initr_trap>:
fff49a18: 94 21 ff f0 stwu r1,-16(r1) fff49a1c: 7c 08 02 a6 mflr r0 fff49a20: 90 01 00 14 stw r0,20(r1) -fff49a24: 80 62 00 44 lwz r3,68(r2) +fff49a24: 80 62 00 38 lwz r3,56(r2) fff49a28: 4b ff 76 19 bl fff41040 <trap_init> fff49a2c: 80 01 00 14 lwz r0,20(r1) fff49a30: 38 60 00 00 li r3,0 fff49a34: 38 21 00 10 addi r1,r1,16 fff49a38: 7c 08 03 a6 mtlr r0
This was causing a consistent hard lockup during the MMC read / loading of the QoriQ FMan firmware on a P2041RDB board.
Re-adding the header causes identical assembly to be emitted and allows the firmware loading and subsequent boot to succeed.
---
arch/powerpc/lib/traps.c | 1 + 1 file changed, 1 insertion(+)
arch/powerpc/lib/traps.c +1 -0
@@ -4,6 +4,7 @@* Wolfgang Denk, DENX Software Engineering, wd at denx.de. */ +#include <common.h> #include <init.h> #include <asm/global_data.h>
Signature · 1 line
2.25.1
PATCH 1/1 powerpc: fix regression in arch_initr_trap()
Hi Matt,
Quoted 15 lines · Matt Merhar
The assembly output of the
arch_initr_trap()function differed by a single byte aftercommon.hwas removed fromtraps.c:
fff49a18<arch_initr_trap>:fff49a18: 94 21 ff f0 stwu r1,-16(r1) fff49a1c: 7c 08 02 a6 mflr r0 fff49a20: 90 01 00 14 stw r0,20(r1)(quoted diff) +1
-1-fff49a24: 80 62 00 44 lwz r3,68(r2) +fff49a24: 80 62 00 38 lwz r3,56(r2) fff49a28: 4b ff 76 19 bl fff41040 <trap_init> fff49a2c: 80 01 00 14 lwz r0,20(r1) fff49a30: 38 60 00 00 li r3,0 fff49a34: 38 21 00 10 addi r1,r1,16 fff49a38: 7c 08 03 a6 mtlr r0
Quoted 8 lines
This was causing a consistent hard lockup during the MMC read / loading of the QoriQ FMan firmware on a P2041RDB board.
Re-adding the header causes identical assembly to be emitted and allows the firmware loading and subsequent boot to succeed.
Fixes 401d1c4f5d ("common: Drop asm/global_data.h from common header")Signed-off-by Matt Merhar <[email protected]>
Did you try to investigate what exactly causes this difference in the assembly code, when the header is not included? Re-adding this header seems like "papering over" the real problem to me.
Thanks,
Stefan
Quoted 11 lines
---
arch/powerpc/lib/traps.c | 1 + 1 file changed, 1 insertion(+)arch/powerpc/lib/traps.c +0
-0
arch/powerpc/lib/traps.c line 10
Quoted 7 lines
arch/powerpc/lib/traps.c +1
-0@@ -4,6 +4,7 @@* Wolfgang Denk, DENX Software Engineering, wd at denx.de. */ +#include <common.h> #include <init.h> #include <asm/global_data.h>
Viele Gr??e,
Stefan
Signature · 3 lines
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-51 Fax: (+49)-8142-66989-80 Email: sr at denx.de
PATCH 1/1 powerpc: fix regression in arch_initr_trap()
Quoted 24 lines · Stefan Roese
Hi Matt,
Quoted 21 lines · Matt Merhar
The assembly output of the
arch_initr_trap()function differed by a single byte aftercommon.hwas removed fromtraps.c:?
fff49a18<arch_initr_trap>:
?fff49a18:????? 94 21 ff f0???? stwu??? r1,-16(r1) ?fff49a1c:????? 7c 08 02 a6???? mflr??? r0 ?fff49a20:????? 90 01 00 14???? stw???? r0,20(r1)(quoted diff) +1
-1-fff49a24:????? 80 62 00 44???? lwz???? r3,68(r2) +fff49a24:????? 80 62 00 38???? lwz???? r3,56(r2)?
fff49a28:????? 4b ff 76 19???? bl?????fff41040<trap_init> ?fff49a2c:????? 80 01 00 14???? lwz???? r0,20(r1) ?fff49a30:????? 38 60 00 00???? li????? r3,0
?fff49a34:????? 38 21 00 10???? addi??? r1,r1,16
?fff49a38:????? 7c 08 03 a6???? mtlr??? r0This was causing a consistent hard lockup during the MMC read / loading of the QoriQ FMan firmware on a P2041RDB board.
Re-adding the header causes identical assembly to be emitted and allows the firmware loading and subsequent boot to succeed.
Quoted 6 lines
Quoted 2 lines
Fixes 401d1c4f5d ("common: Drop asm/global_data.h from common header")Signed-off-by Matt Merhar <[email protected]>Did you try to investigate what exactly causes this difference in the assembly code, when the header is not included? Re-adding this header seems like "papering over" the real problem to me.
Running pahole on traps.o without and with that patch shows (I took P2041RDB_defconfig)
struct global_data {
struct bd_info * bd; /* 0 4 */
long unsigned int flags; /* 4 4 */
unsigned int baudrate; /* 8 4 */
long unsigned int cpu_clk; /* 12 4 */
long unsigned int bus_clk; /* 16 4 */
long unsigned int pci_clk; /* 20 4 */
long unsigned int mem_clk; /* 24 4 */
long unsigned int have_console; /* 28 4 */
long unsigned int env_addr; /* 32 4 */
long unsigned int env_valid; /* 36 4 */
long unsigned int env_has_init; /* 40 4 */
int env_load_prio; /* 44 4 */
long unsigned int ram_base; /* 48 4 *//* XXX 4 bytes hole, try to pack */
phys_addr_t ram_top; /* 56 8 */
/* --- cacheline 1 boundary (64 bytes) --- */
long unsigned int relocaddr; /* 64 4 */
vs
struct global_data {
struct bd_info * bd; /* 0 4 */
long unsigned int flags; /* 4 4 */
unsigned int baudrate; /* 8 4 */
long unsigned int cpu_clk; /* 12 4 */
long unsigned int bus_clk; /* 16 4 */
long unsigned int pci_clk; /* 20 4 */
long unsigned int mem_clk; /* 24 4 */
long unsigned int post_log_word; /* 28 4 */
long unsigned int post_log_res; /* 32 4 */
long unsigned int post_init_f_time; /* 36 4 */
long unsigned int have_console; /* 40 4 */
long unsigned int env_addr; /* 44 4 */
long unsigned int env_valid; /* 48 4 */
long unsigned int env_has_init; /* 52 4 */
int env_load_prio; /* 56 4 */
long unsigned int ram_base; /* 60 4 */ /* --- cacheline 1 boundary (64 bytes) --- */
phys_addr_t ram_top; /* 64 8 */
long unsigned int relocaddr; /* 72 4 */
so it seems to be the
#if defined(CONFIG_POST)
in include/asm-generic/global_data.h which is broken or unreliable or whatever the appropriate way to describe it is.
It would be nice if the linker could catch that kind of thing when collecting debug info from various TUs.
Rasmus
cover sizeof(gd_t) sanity checking
I haven't done a whole lot of testing of these, just enough to see that it normally works _and_ that it would catch the bug Matt reported. I don't have time or resources to find and fix the bugs this would reveal, hence just an RFC for people to consider.
Rasmus Villemoes (2):
build_bug.h: add wrapper for _Static_assert global-data.h: add build-time sanity check of sizeof(struct
global_data)
include/asm-generic/global_data.h | 5 +++++ include/linux/build_bug.h | 19 +++++++++++++++++++ 2 files changed, 24 insertions(+)
Signature · 1 line
2.29.2
RFC PATCH 1/2 build_bug.h: add wrapper for _Static_assert
[Linux commit 6bab69c65013bed5fce9f101a64a84d0385b3946]
BUILD_BUG_ON() is a little annoying, since it cannot be used outside function scope. So one cannot put assertions about the sizeof() a struct next to the struct definition, but has to hide that in some more or less arbitrary function.
Since gcc 4.6 (which is now also the required minimum), there is support for the C11 _Static_assert in all C modes, including gnu89. So add a simple wrapper for that.
_Static_assert() requires a message argument, which is usually quite redundant (and I believe that bug got fixed at least in newer C++ standards), but we can easily work around that with a little macro magic, making it optional.
For example, adding
static_assert(sizeof(struct printf_spec) == 8);
in vsprintf.c and modifying that struct to violate it, one gets
./include/linux/build_bug.h:78:41: error: static assertion failed: "sizeof(struct printf_spec) == 8"
#define __static_assert(expr, msg, ...) _Static_assert(expr, "" msg "")
godbolt.org suggests that _Static_assert() has been support by clang since at least 3.0.0.
--- include/linux/build_bug.h | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
include/linux/build_bug.h +19 -0
@@ -79,6 +79,25 @@*/ #define BUILD_BUG () BUILD_BUG_ON_MSG(1, "BUILD_BUG failed") +/** +* static_assert - check integer constant expression at build time +* +* static_assert() is a wrapper for the C11 _Static_assert, with a +* little macro magic to make the message optional (defaulting to the +* stringification of the tested expression). +* +* Contrary to BUILD_BUG_ON(), static_assert() can be used at global +* scope, but requires the expression to be an integer constant +* expression (i.e., it is not enough that __builtin_constant_p() is +* true for expr). +* +* Also note that BUILD_BUG_ON() fails the build if the condition is +* true, while static_assert() fails the build if the expression is +* false. +*/ +#define static_assert (expr, ...) __static_assert(expr, ##__VA_ARGS__, #expr) +#define __static_assert (expr, msg, ...) _Static_assert(expr, msg) +#endif /* __CHECKER__ */ #endif /* _LINUX_BUILD_BUG_H */
Signature · 1 line
2.29.2
RFC PATCH 1/2 build_bug.h: add wrapper for _Static_assert
Quoted 26 lines · Rasmus Villemoes
[Linux commit
6bab69c65013bed5fce9f101a64a84d0385b3946]
BUILD_BUG_ON()is a little annoying, since it cannot be used outside function scope. So one cannot put assertions about thesizeof()a struct next to the struct definition, but has to hide that in some more or less arbitrary function.Since gcc 4.6 (which is now also the required minimum), there is support for the C11
_Static_assertin all C modes, including gnu89. So add a simple wrapper for that.
_Static_assert()requires a message argument, which is usually quite redundant (and I believe that bug got fixed at least in newer C++ standards), but we can easily work around that with a little macro magic, making it optional.For example, adding
static_assert(sizeof(struct printf_spec) == 8);in
vsprintf.cand modifying that struct to violate it, one gets
./include/linux/build_bug.h:78:41: error: static assertion failed: "sizeof(struct printf_spec)== 8"
#define__static_assert(expr, msg, ...)_Static_assert(expr, "" msg "")
Quoted 7 lines
godbolt.org suggests that
_Static_assert()has been support by clang since at least 3.0.0.Signed-off-by Rasmus Villemoes <[email protected]>--- include/linux/build_bug.h | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
Re: [RFC PATCH 1/2] build_bug.h: add wrapper for _Static_assert
In reply to Rasmus VillemoesQuoted 22 lines · Rasmus Villemoes
[Linux commit
6bab69c65013bed5fce9f101a64a84d0385b3946]
BUILD_BUG_ON()is a little annoying, since it cannot be used outside function scope. So one cannot put assertions about thesizeof()a struct next to the struct definition, but has to hide that in some more or less arbitrary function.Since gcc 4.6 (which is now also the required minimum), there is support for the C11
_Static_assertin all C modes, including gnu89. So add a simple wrapper for that.
_Static_assert()requires a message argument, which is usually quite redundant (and I believe that bug got fixed at least in newer C++ standards), but we can easily work around that with a little macro magic, making it optional.For example, adding
static_assert(sizeof(struct printf_spec) == 8);in
vsprintf.cand modifying that struct to violate it, one gets
Quoted 8 lines
./include/linux/build_bug.h:78:41: error: static assertion failed: "sizeof(struct printf_spec)== 8"
#define__static_assert(expr, msg, ...)_Static_assert(expr, "" msg "")godbolt.org suggests that
_Static_assert()has been support by clang since at least 3.0.0.Signed-off-by Rasmus Villemoes <[email protected]>Reviewed-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
RFC PATCH 2/2 global-data.h: add build-time sanity check of sizeof(struct global_data)
In reply to Rasmus VillemoesThe layout and contents of struct global_data depends on a lot of CONFIG_* preprocessor macros, not all of which are entirely converted to Kconfig - not to mention weird games played here and there. This can result in one translation unit using one definition of struct global_data while the actual layout is another.
That can be very hard to debug. But we already have a mechanism that can help catch such bugs at build time, namely the asm-offsets machinery which is necessary anyway to provide assembly code with the necessary constants. So make sure that every C translation unit that include global_data.h actually sees the same size of struct global_data as that which was seen by the asm-offsets.c TU.
It is likely that this patch will break the build of some boards. For example, without the patch from Matt Merhar
(https://lists.denx.de/pipermail/u-boot/2021-May/450135.html) or some other fix, this breaks P2041RDB_defconfig:
CC arch/powerpc/lib/traps.o
AS arch/powerpc/cpu/mpc85xx/start.o
In file included from include/asm-generic/global_data.h:26,
from ./arch/powerpc/include/asm/global_data.h:109,
from include/init.h:21,
from arch/powerpc/lib/traps.c:7:
include/linux/build_bug.h:99:41: error: static assertion failed: "sizeof(struct global_data) == GD_SIZE"
99 | #define __static_assert(expr, msg, ...) _Static_assert(expr, msg)
| ^~~~~~~~~~~~~~
include/linux/build_bug.h:98:34: note: in expansion of macro ?__static_assert?
98 | #define static_assert(expr, ...) __static_assert(expr, ##__VA_ARGS__, #expr)
| ^~~~~~~~~~~~~~~
include/asm-generic/global_data.h:470:1: note: in expansion of macro ?static_assert?
470 | static_assert(sizeof(struct global_data) == GD_SIZE);
| ^~~~~~~~~~~~~make[1]: *** [scripts/Makefile.build:266: arch/powerpc/lib/traps.o] Error 1
make: *** [Makefile:1753: arch/powerpc/lib] Error 2
make: *** Waiting for unfinished jobs....
--- include/asm-generic/global_data.h | 5 +++++ 1 file changed, 5 insertions(+)
include/asm-generic/global_data.h +5 -0
@@ -23,6 +23,8 @@#include <fdtdec.h> #include <membuff.h> #include <linux/list.h> +#include <linux/build_bug.h> +#include <asm-offsets.h> struct acpi_ctx ;struct driver_rt ;
@@ -464,6 +466,9 @@ struct global_data {char *smbios_version;#endif }; +#ifndef DO_DEPS_ONLY +static_assert (sizeof (struct global_data ) ==GD_SIZE ); +#endif /** * gd_board_type() - retrieve board type
Signature · 1 line
2.29.2
RFC PATCH 2/2 global-data.h: add build-time sanity check of sizeof(struct global_data)
Quoted 34 lines · Rasmus Villemoes
The layout and contents of struct
global_datadepends on a lot of CONFIG_* preprocessor macros, not all of which are entirely converted toKconfig- not to mention weird games played here and there. This can result in one translation unit using one definition of structglobal_datawhile the actual layout is another.That can be very hard to debug. But we already have a mechanism that can help catch such bugs at build time, namely the asm-offsets machinery which is necessary anyway to provide assembly code with the necessary constants. So make sure that every C translation unit that include
global_data.hactually sees the same size of structglobal_dataas that which was seen by theasm-offsets.cTU.It is likely that this patch will break the build of some boards. For example, without the patch from Matt Merhar
(https://lists.denx.de/pipermail/u-boot/2021-May/450135.html) or some other fix, this breaksP2041RDB_defconfig:CC
arch/powerpc/lib/traps.o
ASarch/powerpc/cpu/mpc85xx/start.o
In file included frominclude/asm-generic/global_data.h:26,from ./arch/powerpc/include/asm/global_data.h:109, from include/init.h:21, from arch/powerpc/lib/traps.c:7: include/linux/build_bug.h:99:41: error: static assertion failed: "sizeof(struct global_data) == GD_SIZE" 99 | #define __static_assert(expr, msg, ...) _Static_assert(expr, msg) | ^~~~~~~~~~~~~~ include/linux/build_bug.h:98:34: note: in expansion of macro ?__static_assert? 98 | #define static_assert(expr, ...) __static_assert(expr, ##__VA_ARGS__, #expr) | ^~~~~~~~~~~~~~~ include/asm-generic/global_data.h:470:1: note: in expansion of macro ?static_assert? 470 | static_assert(sizeof(struct global_data) == GD_SIZE); | ^~~~~~~~~~~~~
Quoted 8 lines
make[1]: *** [
scripts/Makefile.build:266:arch/powerpc/lib/traps.o] Error 1
make: *** [Makefile:1753:arch/powerpc/lib] Error 2
make: *** Waiting for unfinished jobs....Signed-off-by Rasmus Villemoes <[email protected]>--- include/asm-generic/global_data.h | 5 +++++ 1 file changed, 5 insertions(+)
Re: [RFC PATCH 2/2] global-data.h: add build-time sanity check of sizeof(struct global_data)
In reply to Rasmus VillemoesQuoted 33 lines · Rasmus Villemoes
The layout and contents of struct
global_datadepends on a lot of CONFIG_* preprocessor macros, not all of which are entirely converted toKconfig- not to mention weird games played here and there. This can result in one translation unit using one definition of structglobal_datawhile the actual layout is another.That can be very hard to debug. But we already have a mechanism that can help catch such bugs at build time, namely the asm-offsets machinery which is necessary anyway to provide assembly code with the necessary constants. So make sure that every C translation unit that include
global_data.hactually sees the same size of structglobal_dataas that which was seen by theasm-offsets.cTU.It is likely that this patch will break the build of some boards. For example, without the patch from Matt Merhar
(https://lists.denx.de/pipermail/u-boot/2021-May/450135.html) or some other fix, this breaksP2041RDB_defconfig:CC
arch/powerpc/lib/traps.o
ASarch/powerpc/cpu/mpc85xx/start.o
In file included frominclude/asm-generic/global_data.h:26,from ./arch/powerpc/include/asm/global_data.h:109, from include/init.h:21, from arch/powerpc/lib/traps.c:7: include/linux/build_bug.h:99:41: error: static assertion failed: "sizeof(struct global_data) == GD_SIZE" 99 | #define __static_assert(expr, msg, ...) _Static_assert(expr, msg) | ^~~~~~~~~~~~~~ include/linux/build_bug.h:98:34: note: in expansion of macro ‘__static_assert’ 98 | #define static_assert(expr, ...) __static_assert(expr, ##__VA_ARGS__, #expr) | ^~~~~~~~~~~~~~~ include/asm-generic/global_data.h:470:1: note: in expansion of macro ‘static_assert’ 470 | static_assert(sizeof(struct global_data) == GD_SIZE); | ^~~~~~~~~~~~~
Quoted 6 lines
make[1]: *** [
scripts/Makefile.build:266:arch/powerpc/lib/traps.o] Error 1
make: *** [Makefile:1753:arch/powerpc/lib] Error 2
make: *** Waiting for unfinished jobs....Signed-off-by Rasmus Villemoes <[email protected]>Reviewed-by Simon Glass <[email protected]>
Applied to u-boot/next, thanks!
Signature · 1 line
Tom
PATCH 1/4 powerpc: Don't use relative include for config.h in global_data.h
In reply to Rasmus VillemoesAs there is an arch/powerpc/include/asm/config.h file using "" to get config.h here can lead to using that rather than include/config.h. This in turn can lead to a mismatch in the size of gd.
--- arch/powerpc/include/asm/global_data.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/include/asm/global_data.h +1 -1
@@ -8,7 +8,7 @@#ifndef __ASM_GBL_DATA_H#define __ASM_GBL_DATA_H -#include "config.h" +#include <config.h> #include "asm/types.h" /* Architecture-specific global data */
Signature · 1 line
2.17.1
PATCH 2/4 Revert "powerpc: fix regression in arch_initr_trap()"
With the changes in commit 588efcdd72fc ("powerpc: Don't use relative include for config.h in global_data.h") fixing the root of the problem, we no longer need this re-inclusion.
This reverts commit f6c0d365d3e8ee8e4fd3ebe2ed957c2bca9d3328.
---
I will need to reword this to have the right commit ID when merging.
--- arch/powerpc/lib/traps.c | 1 - 1 file changed, 1 deletion(-)
arch/powerpc/lib/traps.c +0 -1
@@ -4,7 +4,6 @@* Wolfgang Denk, DENX Software Engineering, [email protected]. */ -#include <common.h> #include <init.h> #include <asm/global_data.h>
Signature · 1 line
2.17.1
Re: [PATCH 2/4] Revert "powerpc: fix regression in arch_initr_trap()"
Quoted 8 lines · Tom Rini
With the changes in commit
588efcdd72fc("powerpc: Don't use relative include forconfig.hinglobal_data.h") fixing the root of the problem, we no longer need this re-inclusion.This reverts commit
f6c0d365d3e8ee8e4fd3ebe2ed957c2bca9d3328.Cc Matt Merhar <[email protected]>Signed-off-by Tom Rini <[email protected]>
Applied to u-boot/master, thanks!
Signature · 1 line
Tom
PATCH 3/4 socfpga64: Do not define CONFIG_SYS_MEM_RESERVE_SECURE to 0
In reply to Tom RiniBased on the comment in socfpga_soc64_common.h, the intention is for CONFIG_SYS_MEM_RESERVE_SECURE to be unused. However, in the code we do:
...
and that will evaluate to true. This leads to unwanted code being compiled. Further, as CONFIG_SYS_MEM_RESERVE_SECURE has not been migrated to Kconfig, this leads to a mismatch in the size of gd depending on if we have or have not also had <configs/BOARD.h> also included yet.
Remove the define as it's not needed.
--- include/configs/socfpga_soc64_common.h | 1 - 1 file changed, 1 deletion(-)
include/configs/socfpga_soc64_common.h +0 -1
@@ -21,7 +21,6 @@/* sysmgr.boot_scratch_cold4 & 5 (64bit) will be used for PSCI_CPU_ON call */ #define CPU_RELEASE_ADDR 0xFFD12210#define CONFIG_SYS_CACHELINE_SIZE 64 -#define CONFIG_SYS_MEM_RESERVE_SECURE 0/* using OCRAM, not DDR */ /* * U-Boot console configurations
Signature · 1 line
2.17.1
Re: [PATCH 3/4] socfpga64: Do not define CONFIG_SYS_MEM_RESERVE_SECURE to 0
Quoted 5 lines · Tom Rini
Based on the comment in
socfpga_soc64_common.h, the intention is forCONFIG_SYS_MEM_RESERVE_SECUREto be unused. However, in the code we do:
...and that will evaluate to true. This leads to unwanted code being
Some cleanup made lines beginning with # disappear? I can _strongly_ recommend setting commit.cleanup to scissors; I have lost count of the number of times I've had a commit message mangled and become practically useless because important lines got removed.
For example
<http://lists.busybox.net/pipermail/busybox/2018-September/086647.html>
became
<https://git.busybox.net/busybox/commit/?id=571e525a141a2de87b9c2ced485745e96418d921>
Rasmus
Re: [PATCH 3/4] socfpga64: Do not define CONFIG_SYS_MEM_RESERVE_SECURE to 0
Quoted 12 lines · Rasmus Villemoes
Quoted 5 lines · Tom Rini
Based on the comment in
socfpga_soc64_common.h, the intention is forCONFIG_SYS_MEM_RESERVE_SECUREto be unused. However, in the code we do:
...and that will evaluate to true. This leads to unwanted code being
Some cleanup made lines beginning with # disappear? I can _strongly_ recommend setting commit.cleanup to scissors; I have lost count of the number of times I've had a commit message mangled and become practically useless because important lines got removed.
Quoted 4 lines
For example
<http://lists.busybox.net/pipermail/busybox/2018-September/086647.html>
became
<https://git.busybox.net/busybox/commit/?id=571e525a141a2de87b9c2ced485745e96418d921>
Bah! Yes, thanks. It should have said
#ifdef CONFIG_SYS_MEM_RESERVE_SECURE
...
#endif
Signature · 1 line
Tom
Re: [PATCH 3/4] socfpga64: Do not define CONFIG_SYS_MEM_RESERVE_SECURE to 0
In reply to Tom RiniQuoted 10 lines · Tom Rini
Based on the comment in
socfpga_soc64_common.h, the intention is forCONFIG_SYS_MEM_RESERVE_SECUREto be unused. However, in the code we do:
...and that will evaluate to true. This leads to unwanted code being compiled. Further, as
CONFIG_SYS_MEM_RESERVE_SECUREhas not been migrated toKconfig, this leads to a mismatch in the size of gd depending on if we have or have not also had <configs/BOARD.h> also included yet.
Quoted 6 lines
Remove the define as it's not needed.
Cc Siew Chin Lim <[email protected]>Cc Chee Hong Ang <[email protected]>Cc Dalon Westergreen <[email protected]>Signed-off-by Tom Rini <[email protected]>
Applied to u-boot/master, thanks!
Signature · 1 line
Tom
PATCH 4/4 global_data: Ensure we have <config.h> when symbols are not in Kconfig yet
In reply to Tom RiniAll symbols that are defined in Kconfig will always be defined (or not) prior to preprocessing due to the -include directive while building. However, symbols which are not yet migrated will only be defined (or not) once the board config.h is included, via <config.h>. While the end goal must be to migrate all symbols, today we have cases where the size of gd will get mismatched within the build, based on include order. Mitigate this by making sure that any <asm/global_data.h> that uses symbols not in Kconfig does start with <config.h>. Remove this when not needed.
--- arch/arc/include/asm/global_data.h | 2 -- arch/arm/include/asm/global_data.h | 2 ++ arch/m68k/include/asm/global_data.h | 2 ++ arch/nds32/include/asm/global_data.h | 2 ++ 4 files changed, 6 insertions(+), 2 deletions(-)
arch/arc/include/asm/global_data.h +0 -2
@@ -6,8 +6,6 @@#ifndef __ASM_ARC_GLOBAL_DATA_H#define __ASM_ARC_GLOBAL_DATA_H -#include <config.h> -#ifndef __ASSEMBLY__/* Architecture-specific global data */ struct arch_global_data {
arch/arm/include/asm/global_data.h +2 -0
@@ -9,6 +9,8 @@#ifndef __ASSEMBLY__ +#include <config.h> +#include <asm/types.h> #include <linux/types.h>
arch/m68k/include/asm/global_data.h +2 -0
@@ -7,6 +7,8 @@ #ifndef __ASM_GBL_DATA_H #define __ASM_GBL_DATA_H +#include <config.h> + /* Architecture-specific global data */ struct arch_global_data { #ifdef CONFIG_SYS_I2C_FSL
arch/nds32/include/asm/global_data.h +2 -0
@@ -17,6 +17,8 @@#ifndef __ASM_GBL_DATA_H#define __ASM_GBL_DATA_H +#include <config.h> +/* Architecture-specific global data */ struct arch_global_data { };
Signature · 1 line
2.17.1
Re: [PATCH 4/4] global_data: Ensure we have <config.h> when symbols are not in Kconfig yet
Quoted 10 lines · Tom Rini
All symbols that are defined in
Kconfigwill always be defined (or not) prior to preprocessing due to the -include directive while building. However, symbols which are not yet migrated will only be defined (or not) once the boardconfig.his included, via <config.h>. While the end goal must be to migrate all symbols, today we have cases where the size of gd will get mismatched within the build, based on include order. Mitigate this by making sure that any <asm/global_data.h> that uses symbols not inKconfigdoes start with <config.h>. Remove this when not needed.
Quoted 6 lines
Cc Alexey Brodkin <[email protected]>Cc Eugeniy Paltsev <[email protected]>Cc Huan Wang <[email protected]>Cc Angelo Dureghello <[email protected]>Cc Rick Chen <[email protected]>Signed-off-by Tom Rini <[email protected]>
Applied to u-boot/master, thanks!
Signature · 1 line
Tom
Re: [PATCH 1/4] powerpc: Don't use relative include for config.h in global_data.h
In reply to Tom RiniQuoted 24 lines · On Thu, 03 Jun 2021 09:38:59 -0400 "Tom Rini" <[email protected]> wrote:
As there is an
arch/powerpc/include/asm/config.hfile using "" to getconfig.hhere can lead to using that rather thaninclude/config.h. This in turn can lead to a mismatch in the size of gd.Cc Matt Merhar <[email protected]>Signed-off-by Tom Rini <[email protected]>--- arch/powerpc/include/asm/global_data.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)(quoted diff) +0
-0
b/arch/powerpc/include/asm/global_data.hindex192a02d347e7..90bf5a2aea5d100644 ---a/arch/powerpc/include/asm/global_data.h+++b/arch/powerpc/include/asm/global_data.h@@ -8,7 +8,7 @@
#ifndef__ASM_GBL_DATA_H
#define__ASM_GBL_DATA_Harch/powerpc/include/asm/global_data.h +1
-2-#include "config.h" +#include <config.h> #include "asm/types.h" /* Architecture-specific global data */ --
Quoted 2 lines
2.17.1
I think this is a better solution, as it also appears to fix another global_data size mismatch caught by the RFC from Rasmus "global-data.h:
add build-time sanity check of sizeof(struct global_data)".
Re: [PATCH 1/4] powerpc: Don't use relative include for config.h in global_data.h
In reply to Tom RiniQuoted 7 lines · Tom Rini
As there is an
arch/powerpc/include/asm/config.hfile using "" to getconfig.hhere can lead to using that rather thaninclude/config.h. This in turn can lead to a mismatch in the size of gd.Cc Matt Merhar <[email protected]>Signed-off-by Tom Rini <[email protected]>Tested-by Matt Merhar <[email protected]>
Applied to u-boot/master, thanks!
Signature · 1 line
Tom
Quoted 15 lines · Matt Merhar
The assembly output of the
arch_initr_trap()function differed by a single byte aftercommon.hwas removed fromtraps.c:
fff49a18<arch_initr_trap>:fff49a18: 94 21 ff f0 stwu r1,-16(r1) fff49a1c: 7c 08 02 a6 mflr r0 fff49a20: 90 01 00 14 stw r0,20(r1)(quoted diff) +1
-1-fff49a24: 80 62 00 44 lwz r3,68(r2) +fff49a24: 80 62 00 38 lwz r3,56(r2) fff49a28: 4b ff 76 19 bl fff41040 <trap_init> fff49a2c: 80 01 00 14 lwz r0,20(r1) fff49a30: 38 60 00 00 li r3,0 fff49a34: 38 21 00 10 addi r1,r1,16 fff49a38: 7c 08 03 a6 mtlr r0
Quoted 8 lines
This was causing a consistent hard lockup during the MMC read / loading of the QoriQ FMan firmware on a P2041RDB board.
Re-adding the header causes identical assembly to be emitted and allows the firmware loading and subsequent boot to succeed.
Fixes 401d1c4f5d ("common: Drop asm/global_data.h from common header")Signed-off-by Matt Merhar <[email protected]>
Applied to u-boot/master, thanks!
Signature · 1 line
Tom