Message ID | 20191015080827.11589-1-fanjinke@hygon.cn |
---|---|
State | Changes Requested |
Headers | show |
Series | [RESEND,RFC,v3] rtc: Fix the AltCentury value on AMD/Hygon platform | expand |
Oh come on, you sent that patch only a week ago and v2 had a so obvious mistake that my trust in your code quality is now very low. On 15/10/2019 16:08:27+0800, Jinke Fan wrote: > When using following operations: > date -s "21190910 19:20:00" > hwclock -w > to change date from 2019 to 2119 for test, it will fail on Hygon > Dhyana and AMD Zen CPUs, while the same operations run ok on Intel i7 > platform. > > MC146818 driver use function mc146818_set_time() to set register > RTC_FREQ_SELECT(RTC_REG_A)'s bit4-bit6 field which means divider stage > reset value on Intel platform to 0x7. > > While AMD/Hygon RTC_REG_A(0Ah)'s bit4 is defined as DV0 [Reference]: > DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1. Bit5-bit6 is defined > as reserved. > > DV0 is set to 1, it will select Bank 1, which will disable AltCentury > register(0x32) access. As UEFI pass acpi_gbl_FADT.century 0x32 > (AltCentury), the CMOS write will be failed on code: > CMOS_WRITE(century, acpi_gbl_FADT.century). > > Correct RTC_REG_A bank select bit(DV0) to 0 on AMD/Hygon CPUs, it will > enable AltCentury(0x32) register writing and finally setup century as > expected. > > Test results on AMD/Hygon machine show that it works as expected. > > Reference: > https://www.amd.com/system/files/TechDocs/51192_Bolton_FCH_RRG.pdf > section: 3.13 Real Time Clock (RTC) > > Reported-by: kbuild test robot <lkp@intel.com> > Signed-off-by: Jinke Fan <fanjinke@hygon.cn> > --- > > v2->v3: > - Make the changes only relevant to AMD/Hygon. > > v1->v2: > - Fix the compile errors on sparc64/alpha platform. > > drivers/rtc/rtc-mc146818-lib.c | 11 ++++++++++- > include/linux/mc146818rtc.h | 6 ++++++ > 2 files changed, 16 insertions(+), 1 deletion(-) > > diff --git a/drivers/rtc/rtc-mc146818-lib.c b/drivers/rtc/rtc-mc146818-lib.c > index 2ecd8752b088..70502881785d 100644 > --- a/drivers/rtc/rtc-mc146818-lib.c > +++ b/drivers/rtc/rtc-mc146818-lib.c > @@ -172,7 +172,16 @@ int mc146818_set_time(struct rtc_time *time) > save_control = CMOS_READ(RTC_CONTROL); > CMOS_WRITE((save_control|RTC_SET), RTC_CONTROL); > save_freq_select = CMOS_READ(RTC_FREQ_SELECT); > - CMOS_WRITE((save_freq_select|RTC_DIV_RESET2), RTC_FREQ_SELECT); > + > +#ifdef CONFIG_X86 > + if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || > + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) > + CMOS_WRITE((save_freq_select & (~RTC_DV0)), RTC_FREQ_SELECT); > + else > + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); > +#else > + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); > +#endif > > #ifdef CONFIG_MACH_DECSTATION > CMOS_WRITE(real_yrs, RTC_DEC_YEAR); > diff --git a/include/linux/mc146818rtc.h b/include/linux/mc146818rtc.h > index 0661af17a758..7066a7bced61 100644 > --- a/include/linux/mc146818rtc.h > +++ b/include/linux/mc146818rtc.h > @@ -86,6 +86,12 @@ struct cmos_rtc_board_info { > /* 2 values for divider stage reset, others for "testing purposes only" */ > # define RTC_DIV_RESET1 0x60 > # define RTC_DIV_RESET2 0x70 > + > +#ifdef CONFIG_X86 > + /* DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1 on AMD/Hygon platform */ > +# define RTC_DV0 0x10 > +#endif > + > /* Periodic intr. / Square wave rate select. 0=none, 1=32.8kHz,... 15=2Hz */ > # define RTC_RATE_SELECT 0x0F > > -- > 2.17.1 >
On 2019/10/15 16:27, Alexandre Belloni wrote: > Oh come on, you sent that patch only a week ago and v2 had a so obvious > mistake that my trust in your code quality is now very low. Hi Alexandre, The patch v3 has been compiled for sparc64 and alpha architectures with: - GCC_VERSION=7.4.0 make.cross ARCH=sparc64 - GCC_VERSION=7.4.0 make.cross ARCH=alpha the result was passed. And tested on Hygon platform, it worked well. As your comment, the modification will be strictly limited to AMD EPYC(17h) and Hygon CPU in the next version. Also I will do more tests on these platforms. Any more suggestions?
On 15/10/2019 16:08:27+0800, Jinke Fan wrote: > When using following operations: > date -s "21190910 19:20:00" > hwclock -w > to change date from 2019 to 2119 for test, it will fail on Hygon > Dhyana and AMD Zen CPUs, while the same operations run ok on Intel i7 > platform. > > MC146818 driver use function mc146818_set_time() to set register > RTC_FREQ_SELECT(RTC_REG_A)'s bit4-bit6 field which means divider stage > reset value on Intel platform to 0x7. > > While AMD/Hygon RTC_REG_A(0Ah)'s bit4 is defined as DV0 [Reference]: > DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1. Bit5-bit6 is defined > as reserved. > > DV0 is set to 1, it will select Bank 1, which will disable AltCentury > register(0x32) access. As UEFI pass acpi_gbl_FADT.century 0x32 > (AltCentury), the CMOS write will be failed on code: > CMOS_WRITE(century, acpi_gbl_FADT.century). > > Correct RTC_REG_A bank select bit(DV0) to 0 on AMD/Hygon CPUs, it will > enable AltCentury(0x32) register writing and finally setup century as > expected. > > Test results on AMD/Hygon machine show that it works as expected. > > Reference: > https://www.amd.com/system/files/TechDocs/51192_Bolton_FCH_RRG.pdf > section: 3.13 Real Time Clock (RTC) > > Reported-by: kbuild test robot <lkp@intel.com> > Signed-off-by: Jinke Fan <fanjinke@hygon.cn> > --- > > v2->v3: > - Make the changes only relevant to AMD/Hygon. > > v1->v2: > - Fix the compile errors on sparc64/alpha platform. > > drivers/rtc/rtc-mc146818-lib.c | 11 ++++++++++- > include/linux/mc146818rtc.h | 6 ++++++ > 2 files changed, 16 insertions(+), 1 deletion(-) > > diff --git a/drivers/rtc/rtc-mc146818-lib.c b/drivers/rtc/rtc-mc146818-lib.c > index 2ecd8752b088..70502881785d 100644 > --- a/drivers/rtc/rtc-mc146818-lib.c > +++ b/drivers/rtc/rtc-mc146818-lib.c > @@ -172,7 +172,16 @@ int mc146818_set_time(struct rtc_time *time) > save_control = CMOS_READ(RTC_CONTROL); > CMOS_WRITE((save_control|RTC_SET), RTC_CONTROL); > save_freq_select = CMOS_READ(RTC_FREQ_SELECT); > - CMOS_WRITE((save_freq_select|RTC_DIV_RESET2), RTC_FREQ_SELECT); > + > +#ifdef CONFIG_X86 > + if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || > + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) > + CMOS_WRITE((save_freq_select & (~RTC_DV0)), RTC_FREQ_SELECT); This should probably use ~RTC_DIV_RESET2. > + else > + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); > +#else > + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); > +#endif Also, later you have: CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT); This may write bit4 again which would make mc146818_get_time fail so you probably want to update save_freq_select. > > #ifdef CONFIG_MACH_DECSTATION > CMOS_WRITE(real_yrs, RTC_DEC_YEAR); > diff --git a/include/linux/mc146818rtc.h b/include/linux/mc146818rtc.h > index 0661af17a758..7066a7bced61 100644 > --- a/include/linux/mc146818rtc.h > +++ b/include/linux/mc146818rtc.h > @@ -86,6 +86,12 @@ struct cmos_rtc_board_info { > /* 2 values for divider stage reset, others for "testing purposes only" */ > # define RTC_DIV_RESET1 0x60 > # define RTC_DIV_RESET2 0x70 > + > +#ifdef CONFIG_X86 > + /* DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1 on AMD/Hygon platform */ > +# define RTC_DV0 0x10 > +#endif > + > /* Periodic intr. / Square wave rate select. 0=none, 1=32.8kHz,... 15=2Hz */ > # define RTC_RATE_SELECT 0x0F > > -- > 2.17.1 >
On 2019/10/20 6:04, Alexandre Belloni wrote: > On 15/10/2019 16:08:27+0800, Jinke Fan wrote: >> save_freq_select = CMOS_READ(RTC_FREQ_SELECT); >> - CMOS_WRITE((save_freq_select|RTC_DIV_RESET2), RTC_FREQ_SELECT); >> + >> +#ifdef CONFIG_X86 >> + if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || >> + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) >> + CMOS_WRITE((save_freq_select & (~RTC_DV0)), RTC_FREQ_SELECT); > > This should probably use ~RTC_DIV_RESET2. Yes, ~RTC_DIV_RESET2 can actually achieve the same effect, because of bit5-bit6 is defined as reserved. >> + else >> + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); >> +#else >> + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); >> +#endif > > Also, later you have: > > CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT); > > This may write bit4 again which would make mc146818_get_time fail so you > probably want to update save_freq_select. Yes, thanks for reminding me.
diff --git a/drivers/rtc/rtc-mc146818-lib.c b/drivers/rtc/rtc-mc146818-lib.c index 2ecd8752b088..70502881785d 100644 --- a/drivers/rtc/rtc-mc146818-lib.c +++ b/drivers/rtc/rtc-mc146818-lib.c @@ -172,7 +172,16 @@ int mc146818_set_time(struct rtc_time *time) save_control = CMOS_READ(RTC_CONTROL); CMOS_WRITE((save_control|RTC_SET), RTC_CONTROL); save_freq_select = CMOS_READ(RTC_FREQ_SELECT); - CMOS_WRITE((save_freq_select|RTC_DIV_RESET2), RTC_FREQ_SELECT); + +#ifdef CONFIG_X86 + if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD || + boot_cpu_data.x86_vendor == X86_VENDOR_HYGON) + CMOS_WRITE((save_freq_select & (~RTC_DV0)), RTC_FREQ_SELECT); + else + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); +#else + CMOS_WRITE((save_freq_select | RTC_DIV_RESET2), RTC_FREQ_SELECT); +#endif #ifdef CONFIG_MACH_DECSTATION CMOS_WRITE(real_yrs, RTC_DEC_YEAR); diff --git a/include/linux/mc146818rtc.h b/include/linux/mc146818rtc.h index 0661af17a758..7066a7bced61 100644 --- a/include/linux/mc146818rtc.h +++ b/include/linux/mc146818rtc.h @@ -86,6 +86,12 @@ struct cmos_rtc_board_info { /* 2 values for divider stage reset, others for "testing purposes only" */ # define RTC_DIV_RESET1 0x60 # define RTC_DIV_RESET2 0x70 + +#ifdef CONFIG_X86 + /* DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1 on AMD/Hygon platform */ +# define RTC_DV0 0x10 +#endif + /* Periodic intr. / Square wave rate select. 0=none, 1=32.8kHz,... 15=2Hz */ # define RTC_RATE_SELECT 0x0F
When using following operations: date -s "21190910 19:20:00" hwclock -w to change date from 2019 to 2119 for test, it will fail on Hygon Dhyana and AMD Zen CPUs, while the same operations run ok on Intel i7 platform. MC146818 driver use function mc146818_set_time() to set register RTC_FREQ_SELECT(RTC_REG_A)'s bit4-bit6 field which means divider stage reset value on Intel platform to 0x7. While AMD/Hygon RTC_REG_A(0Ah)'s bit4 is defined as DV0 [Reference]: DV0 = 0 selects Bank 0, DV0 = 1 selects Bank 1. Bit5-bit6 is defined as reserved. DV0 is set to 1, it will select Bank 1, which will disable AltCentury register(0x32) access. As UEFI pass acpi_gbl_FADT.century 0x32 (AltCentury), the CMOS write will be failed on code: CMOS_WRITE(century, acpi_gbl_FADT.century). Correct RTC_REG_A bank select bit(DV0) to 0 on AMD/Hygon CPUs, it will enable AltCentury(0x32) register writing and finally setup century as expected. Test results on AMD/Hygon machine show that it works as expected. Reference: https://www.amd.com/system/files/TechDocs/51192_Bolton_FCH_RRG.pdf section: 3.13 Real Time Clock (RTC) Reported-by: kbuild test robot <lkp@intel.com> Signed-off-by: Jinke Fan <fanjinke@hygon.cn> --- v2->v3: - Make the changes only relevant to AMD/Hygon. v1->v2: - Fix the compile errors on sparc64/alpha platform. drivers/rtc/rtc-mc146818-lib.c | 11 ++++++++++- include/linux/mc146818rtc.h | 6 ++++++ 2 files changed, 16 insertions(+), 1 deletion(-)