| Message ID | 1291150888-20937-1-git-send-email-galak@kernel.crashing.org |
|---|---|
| State | Accepted |
| Commit | b194577b2995eacbd2a3769269f0af1bc5e22530 |
| Delegated to: | Wolfgang Denk |
| Headers | show |
Dear Kumar Gala, In message <1291150888-20937-1-git-send-email-galak@kernel.crashing.org> you wrote: > Since board_hwconfig & cpu_hwconfig are defined as weak and dont have a > default value they will get put into the BSS if they aren't defined > elsewhere. This is problematic as we try to utilize hwconfig before > we've relocated and thus BSS isn't setup. > > Instead of giving dummy values in the board files that utilize this > feature, we can just initialize the variables to an empty string and > thus move them out of the BSS if they aren't defined elsewhere. > > Also made board_hwconfig & cpu_hwconfig arrays to reduce size associated > with string pointers vs arrays. > > Signed-off-by: Kumar Gala <galak@kernel.crashing.org> > --- > * Removed null checks for board_hwconfig & cpu_hwconfig > > board/freescale/mpc8641hpcn/mpc8641hpcn.c | 3 --- > board/freescale/p2020ds/p2020ds.c | 3 --- > common/hwconfig.c | 14 ++++++-------- > 3 files changed, 6 insertions(+), 14 deletions(-) Applied, thanks. Best regards, Wolfgang Denk
On Tue, 30 Nov 2010 15:01:28 -0600 Kumar Gala <galak@kernel.crashing.org> wrote: > diff --git a/common/hwconfig.c b/common/hwconfig.c > index 3c9759f..da8d3ed 100644 > --- a/common/hwconfig.c > +++ b/common/hwconfig.c > @@ -68,8 +68,8 @@ next: > return NULL; > } > > -const char *cpu_hwconfig __attribute__((weak)); > -const char *board_hwconfig __attribute__((weak)); > +const char cpu_hwconfig[] __attribute__((weak)) = ""; > +const char board_hwconfig[] __attribute__((weak)) = ""; > > #define HWCONFIG_PRE_RELOC_BUF_SIZE 128 > > @@ -96,13 +96,11 @@ static const char *__hwconfig(const char *opt, size_t *arglen) > return hwconfig_parse(env_hwconfig, strlen(env_hwconfig), > opt, ";", ':', arglen); > > - if (board_hwconfig) > - return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), > - opt, ";", ':', arglen); > + return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), > + opt, ";", ':', arglen); > > - if (cpu_hwconfig) > - return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), > - opt, ";", ':', arglen); > + return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), > + opt, ";", ':', arglen); > > return NULL; > } Hmm. "return x; return y; return NULL;" Was the presence of a board hwconfig really intended to override, rather than add to, the cpu hwconfig? Should we check the return of the first hwconfig_parse to see if it found anything? -Scott
On Nov 30, 2010, at 3:12 PM, Scott Wood wrote: > On Tue, 30 Nov 2010 15:01:28 -0600 > Kumar Gala <galak@kernel.crashing.org> wrote: > >> diff --git a/common/hwconfig.c b/common/hwconfig.c >> index 3c9759f..da8d3ed 100644 >> --- a/common/hwconfig.c >> +++ b/common/hwconfig.c >> @@ -68,8 +68,8 @@ next: >> return NULL; >> } >> >> -const char *cpu_hwconfig __attribute__((weak)); >> -const char *board_hwconfig __attribute__((weak)); >> +const char cpu_hwconfig[] __attribute__((weak)) = ""; >> +const char board_hwconfig[] __attribute__((weak)) = ""; >> >> #define HWCONFIG_PRE_RELOC_BUF_SIZE 128 >> >> @@ -96,13 +96,11 @@ static const char *__hwconfig(const char *opt, size_t *arglen) >> return hwconfig_parse(env_hwconfig, strlen(env_hwconfig), >> opt, ";", ':', arglen); >> >> - if (board_hwconfig) >> - return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), >> - opt, ";", ':', arglen); >> + return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), >> + opt, ";", ':', arglen); >> >> - if (cpu_hwconfig) >> - return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), >> - opt, ";", ':', arglen); >> + return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), >> + opt, ";", ':', arglen); >> >> return NULL; >> } > > Hmm. "return x; return y; return NULL;" > > Was the presence of a board hwconfig really intended to override, rather > than add to, the cpu hwconfig? Should we check the return of the first > hwconfig_parse to see if it found anything? Yeah, I'll fix this - didn't even notice it before. I'm going to make it: ret = hwconfig_parse(env_hwconfig, ...) if (ret) return ret; ret = hwconfig_parse(board_hwconfig, ...) if (ret) return ret; return hwconfig_parse(cpu_hwconfig, ...); - k
diff --git a/board/freescale/mpc8641hpcn/mpc8641hpcn.c b/board/freescale/mpc8641hpcn/mpc8641hpcn.c index 812111d..882ff0b 100644 --- a/board/freescale/mpc8641hpcn/mpc8641hpcn.c +++ b/board/freescale/mpc8641hpcn/mpc8641hpcn.c @@ -60,9 +60,6 @@ int checkboard(void) return 0; } -const char *board_hwconfig = "foo:bar=baz"; -const char *cpu_hwconfig = "foo:bar=baz"; - phys_size_t initdram(int board_type) { diff --git a/board/freescale/p2020ds/p2020ds.c b/board/freescale/p2020ds/p2020ds.c index b507677..b05ef98 100644 --- a/board/freescale/p2020ds/p2020ds.c +++ b/board/freescale/p2020ds/p2020ds.c @@ -69,9 +69,6 @@ int checkboard(void) return 0; } -const char *board_hwconfig = "foo:bar=baz"; -const char *cpu_hwconfig = "foo:bar=baz"; - phys_size_t initdram(int board_type) { phys_size_t dram_size = 0; diff --git a/common/hwconfig.c b/common/hwconfig.c index 3c9759f..da8d3ed 100644 --- a/common/hwconfig.c +++ b/common/hwconfig.c @@ -68,8 +68,8 @@ next: return NULL; } -const char *cpu_hwconfig __attribute__((weak)); -const char *board_hwconfig __attribute__((weak)); +const char cpu_hwconfig[] __attribute__((weak)) = ""; +const char board_hwconfig[] __attribute__((weak)) = ""; #define HWCONFIG_PRE_RELOC_BUF_SIZE 128 @@ -96,13 +96,11 @@ static const char *__hwconfig(const char *opt, size_t *arglen) return hwconfig_parse(env_hwconfig, strlen(env_hwconfig), opt, ";", ':', arglen); - if (board_hwconfig) - return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), - opt, ";", ':', arglen); + return hwconfig_parse(board_hwconfig, strlen(board_hwconfig), + opt, ";", ':', arglen); - if (cpu_hwconfig) - return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), - opt, ";", ':', arglen); + return hwconfig_parse(cpu_hwconfig, strlen(cpu_hwconfig), + opt, ";", ':', arglen); return NULL; }
Since board_hwconfig & cpu_hwconfig are defined as weak and dont have a default value they will get put into the BSS if they aren't defined elsewhere. This is problematic as we try to utilize hwconfig before we've relocated and thus BSS isn't setup. Instead of giving dummy values in the board files that utilize this feature, we can just initialize the variables to an empty string and thus move them out of the BSS if they aren't defined elsewhere. Also made board_hwconfig & cpu_hwconfig arrays to reduce size associated with string pointers vs arrays. Signed-off-by: Kumar Gala <galak@kernel.crashing.org> --- * Removed null checks for board_hwconfig & cpu_hwconfig board/freescale/mpc8641hpcn/mpc8641hpcn.c | 3 --- board/freescale/p2020ds/p2020ds.c | 3 --- common/hwconfig.c | 14 ++++++-------- 3 files changed, 6 insertions(+), 14 deletions(-)