Hi,
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...] hwclock: select() to /dev/rtc0 to wait for clock tick timed out
On prior kernels, there is no problem:
# hwclock 2025-10-22 22:21:04.806992-04:00
I reproduced the same failure on 6.18-rc2 and bisected to this commit:
commit 795cda8338eab036013314dbc0b04aae728880ab Author: Esben Haabendal esben@geanix.com Date: Fri May 16 09:23:35 2025 +0200
rtc: interface: Fix long-standing race when setting alarm
This commit was backported to all current 6.x stable branches, as well as 5.15.x, so they all have the same regression.
Reverting this commit on top of 6.18-rc2 corrects the problem.
Let me know if you need any more info!
Thanks, Nick
On Thursday, 23 October 2025 at 06:45, Nick Bowler nbowler@draconx.ca wrote:
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...]
I assume this is 10 seconds long.
hwclock: select() to /dev/rtc0 to wait for clock tick timed out
And this is 100% reproducible, or does it sometimes work and sometimes fail?
On prior kernels, there is no problem:
# hwclock 2025-10-22 22:21:04.806992-04:00
I reproduced the same failure on 6.18-rc2 and bisected to this commit:
commit 795cda8338eab036013314dbc0b04aae728880ab Author: Esben Haabendal esben@geanix.com
Date: Fri May 16 09:23:35 2025 +0200
rtc: interface: Fix long-standing race when setting alarm
This commit was backported to all current 6.x stable branches, as well as 5.15.x, so they all have the same regression.
Reverting this commit on top of 6.18-rc2 corrects the problem.
Let me know if you need any more info!
Are you using the util-linux hwclock command? Which version?
Do you have CONFIG_RTC_INTF_DEV_UIE_EMUL enabled?
Can you run `hwclock --verbose`, both with and without the reverted commit, and send the output from that?
/Esben
On Thu, Oct 23, 2025 at 07:21:21AM +0000, Esben Haabendal wrote:
On Thursday, 23 October 2025 at 06:45, Nick Bowler nbowler@draconx.ca wrote:
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...]
I assume this is 10 seconds long.
Yeah, about that.
hwclock: select() to /dev/rtc0 to wait for clock tick timed out
And this is 100% reproducible, or does it sometimes work and sometimes fail?
It fails every time.
Are you using the util-linux hwclock command? Which version?
hwclock from util-linux 2.40.2
Do you have CONFIG_RTC_INTF_DEV_UIE_EMUL enabled?
No, this option is not enabled.
Can you run `hwclock --verbose`, both with and without the reverted commit, and send the output from that?
6.18-rc2 (broken):
# hwclock --verbose hwclock from util-linux 2.40.2 System Time: 1761226454.799573 Trying to open: /dev/rtc0 Using the rtc interface to the clock. Last drift adjustment done at 1657523820 seconds after 1969 Last calibration done at 1657523820 seconds after 1969 Hardware clock is on UTC time Assuming hardware clock is kept in UTC time. Waiting for clock tick... hwclock: select() to /dev/rtc0 to wait for clock tick timed out ...synchronization failed
6.18-rc2 w/ revert (working):
# hwclock --verbose hwclock from util-linux 2.40.2 System Time: 1761226685.238753 Trying to open: /dev/rtc0 Using the rtc interface to the clock. Last drift adjustment done at 1657523820 seconds after 1969 Last calibration done at 1657523820 seconds after 1969 Hardware clock is on UTC time Assuming hardware clock is kept in UTC time. Waiting for clock tick... ioctl(3, RTC_UIE_ON, 0): Input/output error Waiting in loop for time from /dev/rtc0 to change ...got clock tick Time read from Hardware Clock: 2025/10/23 13:38:06 Hw clock time : 2025/10/23 13:38:06 = 1761226686 seconds since 1969 Time since last adjustment is 103702866 seconds Calculated Hardware Clock drift is 0.000000 seconds 2025-10-23 09:38:05.239100-04:00
Thanks, Nick
Hi, Thorsten here, the Linux kernel's regression tracker. Top-posting for once, to make this easily accessible to everyone.
Just wondering: was this fixed in between? Just asking, as I noticed the culprit was backported to various stable/longterm series recently
Ciao, Thorsten
On 10/23/25 15:39, Nick Bowler wrote:
On Thu, Oct 23, 2025 at 07:21:21AM +0000, Esben Haabendal wrote:
On Thursday, 23 October 2025 at 06:45, Nick Bowler nbowler@draconx.ca wrote:
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...]
I assume this is 10 seconds long.
Yeah, about that.
hwclock: select() to /dev/rtc0 to wait for clock tick timed out
And this is 100% reproducible, or does it sometimes work and sometimes fail?
It fails every time.
Are you using the util-linux hwclock command? Which version?
hwclock from util-linux 2.40.2
Do you have CONFIG_RTC_INTF_DEV_UIE_EMUL enabled?
No, this option is not enabled.
Can you run `hwclock --verbose`, both with and without the reverted commit, and send the output from that?
6.18-rc2 (broken):
# hwclock --verbose hwclock from util-linux 2.40.2 System Time: 1761226454.799573 Trying to open: /dev/rtc0 Using the rtc interface to the clock. Last drift adjustment done at 1657523820 seconds after 1969 Last calibration done at 1657523820 seconds after 1969 Hardware clock is on UTC time Assuming hardware clock is kept in UTC time. Waiting for clock tick... hwclock: select() to /dev/rtc0 to wait for clock tick timed out ...synchronization failed
6.18-rc2 w/ revert (working):
# hwclock --verbose hwclock from util-linux 2.40.2 System Time: 1761226685.238753 Trying to open: /dev/rtc0 Using the rtc interface to the clock. Last drift adjustment done at 1657523820 seconds after 1969 Last calibration done at 1657523820 seconds after 1969 Hardware clock is on UTC time Assuming hardware clock is kept in UTC time. Waiting for clock tick... ioctl(3, RTC_UIE_ON, 0): Input/output error Waiting in loop for time from /dev/rtc0 to change ...got clock tick Time read from Hardware Clock: 2025/10/23 13:38:06 Hw clock time : 2025/10/23 13:38:06 = 1761226686 seconds since 1969 Time since last adjustment is 103702866 seconds Calculated Hardware Clock drift is 0.000000 seconds 2025-10-23 09:38:05.239100-04:00
Thanks, Nick
#regzbot poke
On Thu, Nov 06, 2025 at 11:25:55AM +0100, Thorsten Leemhuis wrote:
Just wondering: was this fixed in between? Just asking, as I noticed the culprit was backported to various stable/longterm series recently
I am not aware of any fix. I just retested both 6.18-rc4 and 6.17.7 and there is no change in behaviour.
Thanks, Nick
On Mon, Dec 08, 2025 at 03:35:42PM +0100, Thorsten Leemhuis wrote:
Lo!
On 11/26/25 04:18, Nick Bowler wrote:
Any thoughts?
Not really, just a vague idea (and reminder, this is not my area or expertise, I'm just tracking regressions):
Two fixes were proposed for the culprit, see:
https://lore.kernel.org/all/BN0PR08MB69510928028C933749F4139383D1A@BN0PR08MB... https://lore.kernel.org/all/BN0PR08MB6951415A751F236375A2945683D1A@BN0PR08MB...
The first link is a patch for a totally different driver, as far as I know not relevant to any system I have, but I guess that makes at least 3 different systems which have regressed...
I can't figure out how to turn the second link into a correctly- formatted patch file, but since it is a one-line change I just manually applied it on top of 6.19-rc1. This appears to fix the problem.
Thanks, Nick
Any thoughts?
The problem is still present in 6.18-rc7 and reverting the commit indicated below still fixes it.
I am also seeing the same failure on a totally different system with Dallas DS1286 RTC, which is also fixed by reverting this commit.
Since the initial report this regression has been further backported to all the remaining longterm kernel series.
Thanks, Nick
On Thu, Oct 23, 2025 at 12:45:13AM -0400, Nick Bowler wrote:
Hi,
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...] hwclock: select() to /dev/rtc0 to wait for clock tick timed out
On prior kernels, there is no problem:
# hwclock 2025-10-22 22:21:04.806992-04:00
I reproduced the same failure on 6.18-rc2 and bisected to this commit:
commit 795cda8338eab036013314dbc0b04aae728880ab Author: Esben Haabendal esben@geanix.com Date: Fri May 16 09:23:35 2025 +0200 rtc: interface: Fix long-standing race when setting alarm
This commit was backported to all current 6.x stable branches, as well as 5.15.x, so they all have the same regression.
Reverting this commit on top of 6.18-rc2 corrects the problem.
Let me know if you need any more info!
Thanks, Nick
Hi Nick,
I have not used the regression tracker before, so let's give it a try:
#regzbot ^introduced: 795cda8338eab036013314dbc0b04aae728880ab
Adrian
On Tue, 2025-11-25 at 22:18 -0500, Nick Bowler wrote:
Any thoughts?
The problem is still present in 6.18-rc7 and reverting the commit indicated below still fixes it.
I am also seeing the same failure on a totally different system with Dallas DS1286 RTC, which is also fixed by reverting this commit.
Since the initial report this regression has been further backported to all the remaining longterm kernel series.
Thanks, Nick
On Thu, Oct 23, 2025 at 12:45:13AM -0400, Nick Bowler wrote:
Hi,
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...] hwclock: select() to /dev/rtc0 to wait for clock tick timed out
On prior kernels, there is no problem:
# hwclock 2025-10-22 22:21:04.806992-04:00
I reproduced the same failure on 6.18-rc2 and bisected to this commit:
commit 795cda8338eab036013314dbc0b04aae728880ab Author: Esben Haabendal esben@geanix.com Date: Fri May 16 09:23:35 2025 +0200 rtc: interface: Fix long-standing race when setting alarm
This commit was backported to all current 6.x stable branches, as well as 5.15.x, so they all have the same regression.
Reverting this commit on top of 6.18-rc2 corrects the problem.
Let me know if you need any more info!
Thanks, Nick
Lo!
On 11/26/25 04:18, Nick Bowler wrote:
Any thoughts?
Not really, just a vague idea (and reminder, this is not my area or expertise, I'm just tracking regressions):
Two fixes were proposed for the culprit, see:
https://lore.kernel.org/all/BN0PR08MB69510928028C933749F4139383D1A@BN0PR08MB... https://lore.kernel.org/all/BN0PR08MB6951415A751F236375A2945683D1A@BN0PR08MB...
Wondering if they might help. Esben might have an idea, but in case Esben does not reply maybe just give them a spin if you have a minute.
Ciao, Thorsten
The problem is still present in 6.18-rc7 and reverting the commit indicated below still fixes it.
I am also seeing the same failure on a totally different system with Dallas DS1286 RTC, which is also fixed by reverting this commit.
Since the initial report this regression has been further backported to all the remaining longterm kernel series.
Thanks, Nick
On Thu, Oct 23, 2025 at 12:45:13AM -0400, Nick Bowler wrote:
Hi,
After a stable kernel update, the hwclock command seems no longer functional on my SPARC system with an ST M48T59Y-70PC1 RTC:
# hwclock [...long delay...] hwclock: select() to /dev/rtc0 to wait for clock tick timed out
On prior kernels, there is no problem:
# hwclock 2025-10-22 22:21:04.806992-04:00
I reproduced the same failure on 6.18-rc2 and bisected to this commit:
commit 795cda8338eab036013314dbc0b04aae728880ab Author: Esben Haabendal esben@geanix.com Date: Fri May 16 09:23:35 2025 +0200 rtc: interface: Fix long-standing race when setting alarm
This commit was backported to all current 6.x stable branches, as well as 5.15.x, so they all have the same regression.
Reverting this commit on top of 6.18-rc2 corrects the problem.
Let me know if you need any more info!
Thanks, Nick
linux-stable-mirror@lists.linaro.org
-
Esben Haabendal -
John Paul Adrian Glaubitz -
Nick Bowler -
Thorsten Leemhuis