| Submitter | Stefan Reinauer |
|---|---|
| Date | 2010-03-29 19:14:44 |
| Message ID | <4BB0FC24.4060401@coresystems.de> |
| Download | mbox | patch |
| Permalink | /patch/1174/ |
| State | New |
| Headers | show |
Comments
On Mon, Mar 29, 2010 at 1:14 PM, Stefan Reinauer <stepan@coresystems.de>wrote: > This patch makes the assumption that the code doing the cmos choice has > CONFIG_USE_OPTION_TABLE enabled. > > Is this always the case? > I thought qemu checked CMOS values to know how much RAM there was, but didn't have USE_OPTION_TABLE set. It's been a long time since I looked at it, though. Thanks, Myles
On 3/29/10 10:01 PM, Myles Watson wrote: > > > On Mon, Mar 29, 2010 at 1:14 PM, Stefan Reinauer > <stepan@coresystems.de <mailto:stepan@coresystems.de>> wrote: > > This patch makes the assumption that the code doing the cmos > choice has CONFIG_USE_OPTION_TABLE enabled. > > Is this always the case? > > I thought qemu checked CMOS values to know how much RAM there was, but > didn't have USE_OPTION_TABLE set. It's been a long time since I > looked at it, though. I thought it would, but it seems it doesn't... If it did, we should leave the cmos primitives always in, and only make the higher level funktions depend on USE_ or HAVE_OPTION_TABLE
Patch
Index: src/include/pc80/mc146818rtc.h =================================================================== --- src/include/pc80/mc146818rtc.h (revision 5314) +++ src/include/pc80/mc146818rtc.h (working copy) @@ -82,7 +82,7 @@ #define PC_CKS_LOC 46 /* coreboot cmos checksum is usually only built over bytes 49..125 */ -#ifdef AUTOCONF_INCLUDED +#if defined(AUTOCONF_INCLUDED) && (CONFIG_USE_OPTION_TABLE == 1) #include <option_table.h> #endif Index: src/pc80/mc146818rtc_early.c =================================================================== --- src/pc80/mc146818rtc_early.c (revision 5314) +++ src/pc80/mc146818rtc_early.c (working copy) @@ -8,6 +8,7 @@ #error "CONFIG_MAX_REBOOT_CNT too high" #endif +#if CONFIG_USE_OPTION_TABLE == 1 static unsigned char cmos_read(unsigned char addr) { int offs = 0; @@ -55,19 +56,23 @@ return sum == old_sum; } +#endif - static inline int last_boot_normal(void) { +#if CONFIG_USE_OPTION_TABLE == 1 unsigned char byte; byte = cmos_read(RTC_BOOT_BYTE); return (byte & (1 << 1)); +#else + return 0; +#endif } static inline int do_normal_boot(void) { unsigned char byte; - +#if CONFIG_USE_OPTION_TABLE == 1 if (cmos_error() || !cmos_chksum_valid()) { /* There are no impossible values, no checksums so just * trust whatever value we have in the the cmos, @@ -105,6 +110,9 @@ cmos_write(byte, RTC_BOOT_BYTE); return (byte & (1<<1)); +#else + return 0; +#endif } static inline unsigned read_option(unsigned start, unsigned size, unsigned def)
This patch makes the assumption that the code doing the cmos choice has CONFIG_USE_OPTION_TABLE enabled. Is this always the case? Signed-off-by: Stefan Reinauer <stepan@coresystems.de>