rtc: backport support for handling dates before 1970
Subsystems (2)
cover rtc: backport support for handling dates before 1970
Hello,
this is a followup to
https://lore.kernel.org/stable/[email protected]
that handled backporting the two patches by Alexandre to the active stable kernels between 6.15 and 5.15. Here comes a backport to 5.10.y, git am handles application to 5.4.y just fine.
Compared to the backport for later kernels I included a major rework of rtc_time64_to_tm() by Cassio Neri. (FTR: I checked, that commit by Cassio Neri isn't the reason we need to fix rtc_time64_to_tm(), the actual problem is older.)
Now that I completed the backport and did some final checks on it I noticed that the problem fixed here is (TTBOMK) a theoretic one because only drivers with .start_secs < 0 are known to have issues and in 5.10 and before there is no such driver. I'm uncertain if this should result in not backporting the changes. I would tend to pick them anyhow, but I won't argue on a veto.
Best regards
Uwe
Alexandre Mergnat (2):
rtc: Make rtc_time64_to_tm() support dates before 1970
rtc: Fix offset calculation for .start_secs < 0
Cassio Neri (1):
rtc: Improve performance of rtc_time64_to_tm(). Add tests.
drivers/rtc/Kconfig | 10 ++++ drivers/rtc/Makefile | 1 + drivers/rtc/class.c | 2 +- drivers/rtc/lib.c | 121 ++++++++++++++++++++++++++++++++--------- drivers/rtc/lib_test.c | 79 +++++++++++++++++++++++++++ 5 files changed, 185 insertions(+), 28 deletions(-) create mode 100644 drivers/rtc/lib_test.c
base-commit: 01e7e36b8606e5d4fddf795938010f7bfa3aa277
Signature · 1 line
2.49.0
PATCH 1/3 rtc: Improve performance of rtc_time64_to_tm(). Add tests.
From: Cassio Neri <[email protected]>
commit 1d1bb12a8b1805ddeef9793ebeb920179fb0fa38 upstream.
The current implementation of rtc_time64_to_tm() contains unnecessary loops, branches and look-up tables. The new one uses an arithmetic-based algorithm appeared in [1] and is approximately 4.3 times faster (YMMV).
The drawback is that the new code isn't intuitive and contains many 'magic numbers' (not unusual for this type of algorithm). However, [1] justifies all those numbers and, given this function's history, the code is unlikely to need much maintenance, if any at all.
Add a KUnit test case that checks every day in a 160,000 years interval starting on 1970-01-01 against the expected result. Add a new config RTC_LIB_KUNIT_TEST symbol to give the option to run this test suite.
[1] Neri, Schneider, "Euclidean Affine Functions and Applications to Calendar Algorithms". https://arxiv.org/abs/2102.06959
--- drivers/rtc/Kconfig | 10 ++++ drivers/rtc/Makefile | 1 + drivers/rtc/lib.c | 103 +++++++++++++++++++++++++++++++---------- drivers/rtc/lib_test.c | 79 +++++++++++++++++++++++++++++++ 4 files changed, 168 insertions(+), 25 deletions(-) create mode 100644 drivers/rtc/lib_test.c
drivers/rtc/Kconfig +10 -0
@@ -10,6 +10,16 @@ config RTC_MC146818_LIB bool select RTC_LIB +config RTC_LIB_KUNIT_TEST + tristate "KUnit test for RTC lib functions" if !KUNIT_ALL_TESTS + depends on KUNIT + default KUNIT_ALL_TESTS + select RTC_LIB + help + Enable this option to test RTC library functions. + + If unsure, say N. + menuconfig RTC_CLASS bool "Real Time Clock" default n
drivers/rtc/Makefile +1 -0
@@ -183,3 +183,4 @@ obj-$(CONFIG_RTC_DRV_WM8350) += rtc-wm8350.o obj-$(CONFIG_RTC_DRV_X1205) += rtc-x1205.o obj-$(CONFIG_RTC_DRV_XGENE) += rtc-xgene.o obj-$(CONFIG_RTC_DRV_ZYNQMP) += rtc-zynqmp.o +obj-$(CONFIG_RTC_LIB_KUNIT_TEST) += lib_test.o
drivers/rtc/lib.c +78 -25
134 lines not on this page · raw message
drivers/rtc/lib_test.c new file +79 -0
79 lines not on this page · raw message
Signature · 1 line
2.49.0
Sasha Levin
maintainerRe: [PATCH 5.10.y 1/3] rtc: Improve performance of rtc_time64_to_tm(). Add tests.
[ Sasha's backport helper bot ]
Hi,
Summary of potential issues:
⚠️ Found follow-up fixes in mainline
The upstream commit SHA1 provided is correct: 1d1bb12a8b1805ddeef9793ebeb920179fb0fa38
WARNING: Author mismatch between patch and upstream commit:
Backport author: <[email protected]> Commit author: Cassio Neri<[email protected]>
Status in newer kernel trees:
6.15.y | Present (exact SHA1) 6.14.y | Present (exact SHA1) 6.12.y | Present (exact SHA1)
6.6.y | Present (exact SHA1)
6.1.y | Present (exact SHA1)
5.15.y | Present (exact SHA1)
Found fixes commits:8a904a3caa88 rtc: test: Fix invalid format specifier.
Note: The patch differs from the upstream commit:
---
1: 1d1bb12a8b180 ! 1: b0856f15968b7 rtc: Improve performance of rtc_time64_to_tm(). Add tests.
@@ Metadata
## Commit message ##
rtc: Improve performance of rtc_time64_to_tm(). Add tests.
+ commit 1d1bb12a8b1805ddeef9793ebeb920179fb0fa38 upstream.
+
The current implementation of rtc_time64_to_tm() contains unnecessary loops, branches and look-up tables. The new one uses an arithmetic-based algorithm appeared in [1] and is approximately 4.3 times faster (YMMV).
@@ Commit message
+ Signed-off-by: Uwe Kleine-König <[email protected]>
## drivers/rtc/Kconfig ##
@@ drivers/rtc/Kconfig: config RTC_MC146818_LIB
---
Results of testing on various branches:
| Branch | Patch Apply | Build Test | |---------------------------|-------------|------------| | stable/linux-5.10.y | Success | Success |
PATCH 2/3 rtc: Make rtc_time64_to_tm() support dates before 1970
In reply to Uwe Kleine-KönigFrom: Alexandre Mergnat <[email protected]>
commit 7df4cfef8b351fec3156160bedfc7d6d29de4cce upstream.
Conversion of dates before 1970 is still relevant today because these dates are reused on some hardwares to store dates bigger than the maximal date that is representable in the device's native format. This prominently and very soon affects the hardware covered by the rtc-mt6397 driver that can only natively store dates in the interval 1900-01-01 up to 2027-12-31. So to store the date 2028-01-01 00:00:00 to such a device, rtc_time64_to_tm() must do the right thing for time=-2208988800.
--- drivers/rtc/lib.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-)
drivers/rtc/lib.c +19 -5
@@ -46,24 +46,38 @@ EXPORT_SYMBOL(rtc_year_days);* rtc_time64_to_tm - converts time64_t to rtc_time. * * @time: The number of seconds since 01-01-1970 00:00:00. -* (Must be positive.) +* Works for values since at least 1900 * @tm: Pointer to the struct rtc_time. */ void rtc_time64_to_tm (time64_t time,struct rtc_time *tm) { -unsigned int secs; -int days; +int days, secs;u64 u64tmp;u32 u32tmp, udays, century, day_of_century, year_of_century, year, day_of_year, month, day;bool is_Jan_or_Feb, is_leap_year; -/* time must be positive */ +/* +* Get days and seconds while preserving the sign to +* handle negative time values (dates before 1970-01-01) +*/ days =div_s64_rem (time,86400 , &secs); +/* +* We need 0 <= secs < 86400 which isn't given for negative +* values of time. Fixup accordingly. +*/ +if (secs <0 ) { + days -=1 ; + secs +=86400 ; + } +/* day of the week, 1970-01-01 was a Thursday */ tm->tm_wday = (days +4 ) %7 ; +/* Ensure tm_wday is always positive */ +if (tm->tm_wday <0 ) + tm->tm_wday +=7 ;/* * The following algorithm is, basically, Proposition 6.3 of Neri
@@ -93,7 +107,7 @@ void rtc_time64_to_tm(time64_t time, struct rtc_time *tm)* thus, is slightly different from [1]. */ - udays = ((u32 ) days) +719468 ; + udays = days +719468 ; u32tmp =4 * udays +3 ; century = u32tmp /146097 ;
Signature · 1 line
2.49.0
Re: [PATCH 5.10.y 2/3] rtc: Make rtc_time64_to_tm() support dates before 1970
Hi all,
Although untested, I'm pretty sure that with very small changes, the previous revision (1d1bb12) can handle dates prior to 1970-01-01 with no need to add extra branches or arithmetic operations. Indeed, 1d1bb12 contains:
<code>
/* time must be positive */
days = div_s64_rem(time, 86400, &secs);
/* day of the week, 1970-01-01 was a Thursday */tm->tm_wday = (days + 4) % 7;
/* long comments */
udays = ((u32) days) + 719468;
</code>
This could have been changed to:
<code>
/* time must be >= -719468 * 86400 which corresponds to 0000-03-01 */ udays = div_u64_rem(time + 719468 * 86400, 86400, &secs);
/* day of the week, 0000-03-01 was a Wednesday (in the proleptic Gregorian calendar) */tm->tm_wday = (days + 3) % 7;
/* long comments */
</code>
Indeed, the addition of 719468 * 86400 to time makes days to be 719468 more than it should be. Therefore, in the calculation of udays, the addition of 719468 becomes unnecessary and thus, udays == days. Moreover, this means that days can be removed altogether and replaced by udays. (Not the other way around because in the remaining code udays must be u32.)
Now, 719468 % 7 = 1 and thus tm->wday is 1 day after what it should be and we correct that by adding 3 instead of 4.
Therefore, I suggest these changes on top of 1d1bb12 instead of those made in 7df4cfe. Since you're working on this, can I please kindly suggest two other changes?
1) Change the reference provided in the long comment. It should say, "The following algorithm is, basically, Figure 12 of Neri and Schneider [1]" and [1] should refer to the published article:
Neri C, Schneider L. Euclidean affine functions and their
application to calendar algorithms. Softw Pract Exper. 2023;53(4):937-970. doi: 10.1002/spe.3172
https://doi.org/10.1002/spe.3172
The article is much better written and clearer than the pre-print currently referred to.
2) Function rtc_time64_to_tm_test_date_range in drivers/rtc/lib_test.c, is a kunit test that checks the result for everyday in a 160000 years range starting at 1970-01-01. It'd be nice if this test is adapted to the new code and starts at 1900-01-01 (technically, it could start at 0000-03-01 but since tm->year counts from 1900, it would be weird to see tm->year == -1900 to mean that the calendar year is 0.) Also 160000 is definitely an overkill (my bad!) and a couple of thousands of years, say 3000, should be more than safe for anyone. :-)
Many thanks,
Cassio.
Quoted 82 lines · Uwe Kleine-König
From: Alexandre Mergnat <[email protected]>
commit
7df4cfef8b351fec3156160bedfc7d6d29de4cceupstream.Conversion of dates before 1970 is still relevant today because these dates are reused on some hardwares to store dates bigger than the maximal date that is representable in the device's native format. This prominently and very soon affects the hardware covered by the rtc-mt6397 driver that can only natively store dates in the interval 1900-01-01 up to 2027-12-31. So to store the date 2028-01-01 00:00:00 to such a device,
rtc_time64_to_tm()must do the right thing for time=-2208988800.Signed-off-by Alexandre Mergnat <[email protected]>Reviewed-by Uwe Kleine-König <[email protected]>Signed-off-by Alexandre Belloni <[email protected]>Signed-off-by Uwe Kleine-König <[email protected]>--- drivers/rtc/lib.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-)drivers/rtc/lib.c +19
-6@@ -46,24 +46,38 @@ EXPORT_SYMBOL(rtc_year_days);* rtc_time64_to_tm - converts time64_t to rtc_time. * * @time: The number of seconds since 01-01-1970 00:00:00. -* (Must be positive.) +* Works for values since at least 1900 * @tm: Pointer to the struct rtc_time. */ void rtc_time64_to_tm (time64_t time,struct rtc_time *tm) { -unsigned int secs; -int days; +int days, secs;u64 u64tmp;u32 u32tmp, udays, century, day_of_century, year_of_century, year, day_of_year, month, day;bool is_Jan_or_Feb, is_leap_year; -/* time must be positive */ +/* +* Get days and seconds while preserving the sign to +* handle negative time values (dates before 1970-01-01) +*/ days =div_s64_rem (time,86400 , &secs); +/* +* We need 0 <= secs < 86400 which isn't given for negative +* values of time. Fixup accordingly. +*/ +if (secs <0 ) { + days -=1 ; + secs +=86400 ; + } +/* day of the week, 1970-01-01 was a Thursday */ tm->tm_wday = (days +4 ) %7 ; +/* Ensure tm_wday is always positive */ +if (tm->tm_wday <0 ) + tm->tm_wday +=7 ;/* * The following algorithm is, basically, Proposition 6.3 of Neri @@ -93,7 +107,7 @@ void rtc_time64_to_tm(time64_t time, struct rtc_time *tm)* thus, is slightly different from [1]. */ - udays = ((u32 ) days) +719468 ; + udays = days +719468 ; u32tmp =4 * udays +3 ; century = u32tmp /146097 ; --2.49.0
A message that is not in the archive (sent off-list or not yet received): CAOfgUPg0Z6e5+awuqVMa7QUPiJ7aPp-dX6QNk80Y-bhpBYcsoQ@mail.gmail.com
Re: [PATCH 5.10.y 2/3] rtc: Make rtc_time64_to_tm() support dates before 1970
Hello Cassio,
thanks for your input.
Quoted 37 lines · Cassio Neri
Although untested, I'm pretty sure that with very small changes, the previous revision (1d1bb12) can handle dates prior to 1970-01-01 with no need to add extra branches or arithmetic operations. Indeed, 1d1bb12 contains:
<code>
/* time must be positive */
days =div_s64_rem(time, 86400, &secs);/* day of the week, 1970-01-01 was a Thursday */
tm->tm_wday= (days + 4) % 7;/* long comments */
udays = ((u32) days) + 719468;
</code>This could have been changed to:
<code>
/* time must be >= -719468 * 86400 which corresponds to 0000-03-01 */ udays =div_u64_rem(time + 719468 * 86400, 86400, &secs);/* day of the week, 0000-03-01 was a Wednesday (in the proleptic Gregorian calendar) */
tm->tm_wday= (days + 3) % 7;/* long comments */
</code>Indeed, the addition of 719468 * 86400 to
timemakesdaysto be 719468 more than it should be. Therefore, in the calculation ofudays, the addition of 719468 becomes unnecessary and thus,udays == days. Moreover, this means thatdayscan be removed altogether and replaced byudays. (Not the other way around because in the remaining codeudaysmust be u32.)
Quoted 6 lines
Now, 719468 % 7 = 1 and thus
tm->wdayis 1 day after what it should be and we correct that by adding 3 instead of 4.Therefore, I suggest these changes on top of 1d1bb12 instead of those made in 7df4cfe. Since you're working on this, can I please kindly suggest two other changes?
It's to late for "instead", and we're discussing a backport to stable for a commit that is already in v6.16-rc1.
While your concerns are correct (though I didn't check the details yet), I claim that 7df4cfef8b35 is correct and it's the right thing to backport that today. Incremental changes can then go in the development version (and backported if deemed necessary).
Quoted 11 lines
1) Change the reference provided in the long comment. It should say, "The following algorithm is, basically, Figure 12 of Neri and Schneider [1]" and
[1] should refer to the published article:Neri C, Schneider L. Euclidean affine functions and their application to
calendar algorithms. Softw Pract Exper. 2023;53(4):937-970. doi:
10.1002/spe.3172
https://doi.org/10.1002/spe.3172The article is much better written and clearer than the pre-print currently referred to.
I'll add that to my todo list. (that = improving rtc_time64_to_tm() and reading your paper :-)
Quoted 8 lines
2) Function
rtc_time64_to_tm_test_date_rangeindrivers/rtc/lib_test.c, is a kunit test that checks the result for everyday in a 160000 years range starting at 1970-01-01. It'd be nice if this test is adapted to the new code and starts at 1900-01-01 (technically, it could start at 0000-03-01 but sincetm->yearcounts from 1900, it would be weird to seetm->year== -1900 to mean that the calendar year is 0.) Also 160000 is definitely an overkill (my bad!) and a couple of thousands of years, say 3000, should be more than safe for anyone. :-)
I already did 2), see https://git.kernel.org/linus/ccb2dba3c19f.
Best regards
Uwe
Alexandre Belloni
maintainerRe: [PATCH 5.10.y 2/3] rtc: Make rtc_time64_to_tm() support dates before 1970
In reply to a message not in the archiveHello Cassio,
Quoted 50 lines · Cassio Neri
Hi all,
Although untested, I'm pretty sure that with very small changes, the previous revision (1d1bb12) can handle dates prior to 1970-01-01 with no need to add extra branches or arithmetic operations. Indeed, 1d1bb12 contains:
<code>
/* time must be positive */
days =div_s64_rem(time, 86400, &secs);/* day of the week, 1970-01-01 was a Thursday */
tm->tm_wday= (days + 4) % 7;/* long comments */
udays = ((u32) days) + 719468;
</code>This could have been changed to:
<code>
/* time must be >= -719468 * 86400 which corresponds to 0000-03-01 */ udays =div_u64_rem(time + 719468 * 86400, 86400, &secs);/* day of the week, 0000-03-01 was a Wednesday (in the proleptic Gregorian calendar) */
tm->tm_wday= (days + 3) % 7;/* long comments */
</code>Indeed, the addition of 719468 * 86400 to
timemakesdaysto be 719468 more than it should be. Therefore, in the calculation ofudays, the addition of 719468 becomes unnecessary and thus,udays == days. Moreover, this means thatdayscan be removed altogether and replaced byudays. (Not the other way around because in the remaining codeudaysmust be u32.)Now, 719468 % 7 = 1 and thus
tm->wdayis 1 day after what it should be and we correct that by adding 3 instead of 4.Therefore, I suggest these changes on top of 1d1bb12 instead of those made in 7df4cfe. Since you're working on this, can I please kindly suggest two other changes?
1) Change the reference provided in the long comment. It should say, "The following algorithm is, basically, Figure 12 of Neri and Schneider [1]" and
[1] should refer to the published article:
Quoted 8 lines
Neri C, Schneider L. Euclidean affine functions and their application to
calendar algorithms. Softw Pract Exper. 2023;53(4):937-970. doi:
10.1002/spe.3172
https://doi.org/10.1002/spe.3172The article is much better written and clearer than the pre-print currently referred to.
Thanks for your input, I wanted to look again at your paper and make those optimizations which is why I took so long to review the original patch. Unfortunately, I didn't have the time before the merge window.
I would also gladly take patches for this if you are up for the task.
Quoted 8 lines
2) Function
rtc_time64_to_tm_test_date_rangeindrivers/rtc/lib_test.c, is a kunit test that checks the result for everyday in a 160000 years range starting at 1970-01-01. It'd be nice if this test is adapted to the new code and starts at 1900-01-01 (technically, it could start at 0000-03-01 but sincetm->yearcounts from 1900, it would be weird to seetm->year== -1900 to mean that the calendar year is 0.) Also 160000 is definitely an overkill (my bad!) and a couple of thousands of years, say 3000, should be more than safe for anyone. :-)
This is also something on my radar as some have been complaining about the time it takes to run those tests.
Quoted 93 lines
Many thanks,
Cassio.Quoted 84 lines · Uwe Kleine-König
From: Alexandre Mergnat <[email protected]>
commit
7df4cfef8b351fec3156160bedfc7d6d29de4cceupstream.Conversion of dates before 1970 is still relevant today because these dates are reused on some hardwares to store dates bigger than the maximal date that is representable in the device's native format. This prominently and very soon affects the hardware covered by the rtc-mt6397 driver that can only natively store dates in the interval 1900-01-01 up to 2027-12-31. So to store the date 2028-01-01 00:00:00 to such a device,
rtc_time64_to_tm()must do the right thing for time=-2208988800.Signed-off-by Alexandre Mergnat <[email protected]>Reviewed-by Uwe Kleine-König <[email protected]>Link:
https://lore.kernel.org/r/[email protected]Signed-off-by Alexandre Belloni <[email protected]>Signed-off-by Uwe Kleine-König <[email protected]>--- drivers/rtc/lib.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-)drivers/rtc/lib.c +19
-5@@ -46,24 +46,38 @@ EXPORT_SYMBOL(rtc_year_days);* rtc_time64_to_tm - converts time64_t to rtc_time. * * @time: The number of seconds since 01-01-1970 00:00:00. -* (Must be positive.) +* Works for values since at least 1900 * @tm: Pointer to the struct rtc_time. */ void rtc_time64_to_tm (time64_t time,struct rtc_time *tm) { -unsigned int secs; -int days; +int days, secs;u64 u64tmp;u32 u32tmp, udays, century, day_of_century, year_of_century, year, day_of_year, month, day;bool is_Jan_or_Feb, is_leap_year; -/* time must be positive */ +/* +* Get days and seconds while preserving the sign to +* handle negative time values (dates before 1970-01-01) +*/ days =div_s64_rem (time,86400 , &secs); +/* +* We need 0 <= secs < 86400 which isn't given for negative +* values of time. Fixup accordingly. +*/ +if (secs <0 ) { + days -=1 ; + secs +=86400 ; + } +/* day of the week, 1970-01-01 was a Thursday */ tm->tm_wday = (days +4 ) %7 ; +/* Ensure tm_wday is always positive */ +if (tm->tm_wday <0 ) + tm->tm_wday +=7 ;/* * The following algorithm is, basically, Proposition 6.3 of Neri @@ -93,7 +107,7 @@ void rtc_time64_to_tm(time64_t time, struct rtc_time*tm) * thus, is slightly different from [1]. */ - udays = ((u32 ) days) +719468 ; + udays = days +719468 ; u32tmp =4 * udays +3 ;century = u32tmp / 146097;--
2.49.0
Re: [PATCH 5.10.y 2/3] rtc: Make rtc_time64_to_tm() support dates before 1970
Hi Uwe and Alexandre,
Thank both of you for your replies.
Sure, I understand that you couldn't miss the merge window and that improvements can be made later. I wish I could contribute with a patch but I don't have much bandwidth now. I've obviously done this in the past but I have already forgotten all the processes and figuring that out again would be a steep learning curve for me. In any case, please do not hesitate to get in touch if you think I can help.
I'll be happy if you read my paper but here is another source which gives a very good overview and, I dare to say, is more entertaining :-)
https://www.youtube.com/watch?v=J9KijLyP-yg
(The talk is about the algorithms and has close to nothing to do with C++.)
Thank you again and best wishes,
Cassio.
Quoted 174 lines · Alexandre Belloni
Hello Cassio,
Quoted 50 lines · Cassio Neri
Hi all,
Although untested, I'm pretty sure that with very small changes, the previous revision (1d1bb12) can handle dates prior to 1970-01-01 with no need to add extra branches or arithmetic operations. Indeed, 1d1bb12 contains:
<code>
/* time must be positive */
days =div_s64_rem(time, 86400, &secs);/* day of the week, 1970-01-01 was a Thursday */
tm->tm_wday= (days + 4) % 7;/* long comments */
udays = ((u32) days) + 719468;
</code>This could have been changed to:
<code>
/* time must be >= -719468 * 86400 which corresponds to 0000-03-01 */ udays =div_u64_rem(time + 719468 * 86400, 86400, &secs);/* day of the week, 0000-03-01 was a Wednesday (in the proleptic Gregorian calendar) */
tm->tm_wday= (days + 3) % 7;/* long comments */
</code>Indeed, the addition of 719468 * 86400 to
timemakesdaysto be 719468 more than it should be. Therefore, in the calculation ofudays, the addition of 719468 becomes unnecessary and thus,udays == days. Moreover, this means thatdayscan be removed altogether and replaced byudays. (Not the other way around because in the remaining codeudaysmust be u32.)Now, 719468 % 7 = 1 and thus
tm->wdayis 1 day after what it should be and we correct that by adding 3 instead of 4.Therefore, I suggest these changes on top of 1d1bb12 instead of those made in 7df4cfe. Since you're working on this, can I please kindly suggest two other changes?
1) Change the reference provided in the long comment. It should say, "The following algorithm is, basically, Figure 12 of Neri and Schneider [1]" and
[1] should refer to the published article:Quoted 8 lines
Neri C, Schneider L. Euclidean affine functions and their application to
calendar algorithms. Softw Pract Exper. 2023;53(4):937-970. doi:
10.1002/spe.3172
https://doi.org/10.1002/spe.3172The article is much better written and clearer than the pre-print currently referred to.
Thanks for your input, I wanted to look again at your paper and make those optimizations which is why I took so long to review the original patch. Unfortunately, I didn't have the time before the merge window.
I would also gladly take patches for this if you are up for the task.
Quoted 8 lines
2) Function
rtc_time64_to_tm_test_date_rangeindrivers/rtc/lib_test.c, is a kunit test that checks the result for everyday in a 160000 years range starting at 1970-01-01. It'd be nice if this test is adapted to the new code and starts at 1900-01-01 (technically, it could start at 0000-03-01 but sincetm->yearcounts from 1900, it would be weird to seetm->year== -1900 to mean that the calendar year is 0.) Also 160000 is definitely an overkill (my bad!) and a couple of thousands of years, say 3000, should be more than safe for anyone. :-)This is also something on my radar as some have been complaining about the time it takes to run those tests.
Quoted 93 lines
Many thanks,
Cassio.Quoted 84 lines · Uwe Kleine-König
From: Alexandre Mergnat <[email protected]>
commit
7df4cfef8b351fec3156160bedfc7d6d29de4cceupstream.Conversion of dates before 1970 is still relevant today because these dates are reused on some hardwares to store dates bigger than the maximal date that is representable in the device's native format. This prominently and very soon affects the hardware covered by the rtc-mt6397 driver that can only natively store dates in the interval 1900-01-01 up to 2027-12-31. So to store the date 2028-01-01 00:00:00 to such a device,
rtc_time64_to_tm()must do the right thing for time=-2208988800.Signed-off-by Alexandre Mergnat <[email protected]>Reviewed-by Uwe Kleine-König <[email protected]>Link:
https://lore.kernel.org/r/[email protected]Signed-off-by Alexandre Belloni <[email protected]>Signed-off-by Uwe Kleine-König <[email protected]>--- drivers/rtc/lib.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-)drivers/rtc/lib.c +19
-5@@ -46,24 +46,38 @@ EXPORT_SYMBOL(rtc_year_days);* rtc_time64_to_tm - converts time64_t to rtc_time. * * @time: The number of seconds since 01-01-1970 00:00:00. -* (Must be positive.) +* Works for values since at least 1900 * @tm: Pointer to the struct rtc_time. */ void rtc_time64_to_tm (time64_t time,struct rtc_time *tm) { -unsigned int secs; -int days; +int days, secs;u64 u64tmp;u32 u32tmp, udays, century, day_of_century, year_of_century, year, day_of_year, month, day;bool is_Jan_or_Feb, is_leap_year; -/* time must be positive */ +/* +* Get days and seconds while preserving the sign to +* handle negative time values (dates before 1970-01-01) +*/ days =div_s64_rem (time,86400 , &secs); +/* +* We need 0 <= secs < 86400 which isn't given for negative +* values of time. Fixup accordingly. +*/ +if (secs <0 ) { + days -=1 ; + secs +=86400 ; + } +/* day of the week, 1970-01-01 was a Thursday */ tm->tm_wday = (days +4 ) %7 ; +/* Ensure tm_wday is always positive */ +if (tm->tm_wday <0 ) + tm->tm_wday +=7 ;/* * The following algorithm is, basically, Proposition 6.3 of Neri @@ -93,7 +107,7 @@ void rtc_time64_to_tm(time64_t time, struct rtc_time*tm) * thus, is slightly different from [1]. */ - udays = ((u32 ) days) +719468 ; + udays = days +719468 ; u32tmp =4 * udays +3 ;century = u32tmp / 146097;--
2.49.0
Re: [PATCH 5.10.y 2/3] rtc: Make rtc_time64_to_tm() support dates before 1970
In reply to Uwe Kleine-König[ Sasha's backport helper bot ]
Hi,
✅ All tests passed successfully. No issues detected. No action required from the submitter.
The upstream commit SHA1 provided is correct: 7df4cfef8b351fec3156160bedfc7d6d29de4cce
WARNING: Author mismatch between patch and upstream commit:
Backport author: <[email protected]> Commit author: Alexandre Mergnat<[email protected]>
Status in newer kernel trees:
6.15.y | Present (different SHA1: aa9745977724) 6.14.y | Present (different SHA1: 1fd85fd8b7ba) 6.12.y | Present (different SHA1: a6a55fe660f8)
6.6.y | Present (different SHA1: 9af5a32330f8) 6.1.y | Present (different SHA1: 75cd3ea1caef) 5.15.y | Present (different SHA1: fc02a9de9a12)
Note: The patch differs from the upstream commit:
---
1: 7df4cfef8b351 ! 1: c046d6497bd0a rtc: Make rtc_time64_to_tm() support dates before 1970
@@ Metadata
## Commit message ##
rtc: Make rtc_time64_to_tm() support dates before 1970
+ commit 7df4cfef8b351fec3156160bedfc7d6d29de4cce upstream.
+
Conversion of dates before 1970 is still relevant today because these dates are reused on some hardwares to store dates bigger than the maximal date that is representable in the device's native format.
@@ Commit message
+ Signed-off-by: Uwe Kleine-König <[email protected]>
## drivers/rtc/lib.c ##
@@ drivers/rtc/lib.c: EXPORT_SYMBOL(rtc_year_days);
---
Results of testing on various branches:
| Branch | Patch Apply | Build Test | |---------------------------|-------------|------------| | stable/linux-5.15.y | Success | Success |
PATCH 3/3 rtc: Fix offset calculation for .start_secs < 0
In reply to Uwe Kleine-KönigFrom: Alexandre Mergnat <[email protected]>
commit fe9f5f96cfe8b82d0f24cbfa93718925560f4f8d upstream.
The comparison
rtc->start_secs > rtc->range_max
has a signed left-hand side and an unsigned right-hand side. So the comparison might become true for negative start_secs which is interpreted as a (possibly very large) positive value.
As a negative value can never be bigger than an unsigned value the correct representation of the (mathematical) comparison
rtc->start_secs > rtc->range_max
in C is:
rtc->start_secs >= 0 && rtc->start_secs > rtc->range_max
Use that to fix the offset calculation currently used in the rtc-mt6397 driver.
--- drivers/rtc/class.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/rtc/class.c +1 -1
@@ -314,7 +314,7 @@ static void rtc_device_get_offset(struct rtc_device *rtc)* * Otherwise the offset seconds should be 0. */ -if (rtc->start_secs > rtc->range_max || +if ((rtc->start_secs >=0 && rtc->start_secs > rtc->range_max) || rtc->start_secs + range_secs -1 < rtc->range_min) rtc->offset_secs = rtc->start_secs - rtc->range_min;else if (rtc->start_secs > rtc->range_min)
Signature · 1 line
2.49.0
Re: [PATCH 5.10.y 3/3] rtc: Fix offset calculation for .start_secs < 0
[ Sasha's backport helper bot ]
Hi,
✅ All tests passed successfully. No issues detected. No action required from the submitter.
The upstream commit SHA1 provided is correct: fe9f5f96cfe8b82d0f24cbfa93718925560f4f8d
WARNING: Author mismatch between patch and upstream commit:
Backport author: <[email protected]> Commit author: Alexandre Mergnat<[email protected]>
Status in newer kernel trees:
6.15.y | Present (different SHA1: cea549ca705b) 6.14.y | Present (different SHA1: 57082e0300d5) 6.12.y | Present (different SHA1: 6b482b16f32e)
6.6.y | Present (different SHA1: 3d8efdcbc6ce) 6.1.y | Present (different SHA1: b2677da58ed4) 5.15.y | Present (different SHA1: 0931dc7be5dd)
Note: The patch differs from the upstream commit:
---
1: fe9f5f96cfe8b ! 1: 8aa260951391e rtc: Fix offset calculation for .start_secs < 0
@@ Metadata
## Commit message ##
rtc: Fix offset calculation for .start_secs < 0
+ commit fe9f5f96cfe8b82d0f24cbfa93718925560f4f8d upstream.
+
The comparison
rtc->start_secs > rtc->range_max
@@ Commit message
+ Signed-off-by: Uwe Kleine-König <[email protected]>
## drivers/rtc/class.c ##
@@ drivers/rtc/class.c: static void rtc_device_get_offset(struct rtc_device *rtc)
---
Results of testing on various branches:
| Branch | Patch Apply | Build Test | |---------------------------|-------------|------------| | stable/linux-5.15.y | Success | Success |