2.6.25.2 - Jiffies/Time jumping back and forth (Regereesion from 2.6.24)
----- Original Message ----
Quoted 9 lines
From: john stultz <[email protected]>
To: Martin Knoblauch <[email protected]>
Cc: [email protected]
Sent: Wednesday, May 7, 2008 7:07:49 PM
Subject: Re: 2.6.25.2 - Jiffies/Time jumping back and forth (Regereesion from 2.6.24)Quoted 2 lines · Martin Knoblauch
Hi,
Quoted 9 lines
Quoted 1 line
any news on the one below? I upgraded to 2.6.25.2 and the problem still
persists. Would adding a bugzilla actually help ? :-)
With this issue you're seeing is only the printk timestamps jumping back, correct?
correct.
Quoted 4 lines
For example, you're not seeing the output from
gettimeofday()or the actual jiffies value (as seen from maybe a driver) jump backward? Is that right?
correct again. As far as I can see, other time related things seem to be OK, including gettimeofday.
Quoted 5 lines
The printk timestamps may jump around a bit at bootup, especially if you have unsynced TSCs, so that is a known issue. However its not known to have any actual negative effect other then maybe causing the dmesg log to be a bit confusing.
the problem, is not just at bootup and the jumping. During the *complete* lifetime of the kernel I see:
a) the dmesg timestamps jumping back and forth
b) the values are much to small and have absolutely no relation to "seconds after bootup", as they have with 2.6.24 and before.
But you are right, it might not be a jiffie problem, but a "dmesg timestamp" thing.
Cheers
Martin
Quoted 9 lines · Martin Knoblauch
Quoted 5 lines
The printk timestamps may jump around a bit at bootup, especially if you have unsynced TSCs, so that is a known issue. However its not known to have any actual negative effect other then maybe causing the dmesg log to be a bit confusing.
the problem, is not just at bootup and the jumping. During the
*complete* lifetime of the kernel I see:
Quoted 6 lines
a) the dmesg timestamps jumping back and forth
b) the values are much to small and have absolutely no relation to "seconds after bootup", as they have with 2.6.24 and before.But you are right, it might not be a jiffie problem, but a "dmesg
timestamp" thing.
Hrmm.. So the issue persisting over the lifetime of the kernel is a new one by me. This may also have some effect in scheduling decisions (not necessarily causing a correctness issue, but a performance one). Also it may have some effect on thread cputime values.
Please do open a bug and provide example dmesg output. I'll try to tag some folks who have access to the same system type.
thanks
-john