x86: Reduce verbosity of "PAT enabled" kernel message
PATCH 2/2 x86: Reduce verbosity of "PAT enabled" kernel message
On modern systems, the kernel prints the message
x86 PAT enabled: cpu 0, old 0x7040600070406, new 0x7010600070106
once for every CPU. This gets kind of ridiculous on huge systems; for example, on a 64-thread system I was lucky enough to get:
dmesg| grep 'PAT enabled' | wc
64 704 5174
There is already a BUG() if non-boot CPUs have PAT capabilities that don't match the boot CPU, so just print the message on the boot CPU. (I kept the print after the wrmsrl() that enables PAT, so that the log output continues to mean that the system survived enabling PAT on the boot CPU)
--- arch/x86/mm/pat.c | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-)
arch/x86/mm/pat.c +5 -2
@@ -81,6 +81,7 @@ enum {void pat_init (void ) {u64 pat; +bool boot_cpu = !boot_pat_state;if (!pat_enabled)return ;
@@ -122,8 +123,10 @@ void pat_init(void)rdmsrl (MSR_IA32_CR_PAT , boot_pat_state);wrmsrl (MSR_IA32_CR_PAT , pat); -printk (KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n" , -smp_processor_id (), boot_pat_state, pat); + +if (boot_cpu) +printk (KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n" , +smp_processor_id (), boot_pat_state, pat); }#undef PAT
By the way, there is further work to be done in this direction... these were the easy ones (that save me 127 lines of dmesg :).
The philosophy should probably be things that we would want to see to debug crashes at boot should go into the kernel log -- that makes it easy for users to grab the output and send it on for debugging. If we can move other stuff somewhere that users can grab easily after boot, then that's probably worth it.
I think the next things to look at are:
* secondary CPU booting:
Calibrating delay using timer specific routine.. xxxx BogoMIPS (lpj=xxxx)
CPU: Physical Processor ID: 2
CPU: Processor Core ID: 0
CPU: L1 I cache: xxxx, L1 D cache: xxxx
CPU: L2 cache: xxxx
CPU: L3 cache: xxxx
CPU 1/0x40 -> Node 0
mce: CPU supports xxxx MCE banks
CPU1: Thermal monitoring enabled (TM1)
CPU 1 MCA banks ....
where all 64 CPUs are exactly the same -- I guess checking for non-homogenous systems might be kind of complicated but we really shouldn't print all this for every CPU when they all match the boot CPU precisely.
* scheduler domain stuff:
CPU0 attaching sched-domain:
domain 0: span 0,32 level SIBLING
groups: 0 (cpu_power = xxxx) 32 (cpu_power = xxxx)
domain 1: span 0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60 level MC
groups: 0,32 (cpu_power = xxxx) 4,36 (cpu_power = xxxx) 8,40 (cpu_power = xxxx) 12,44 (cpu_power = xxxx) 16,48 (cpu_power = xxxx) 20,52 (cpu_power = xxxx) 24,56 (cpu_power = xxxx) 28,60 (cpu_power = xxxx)
domain 2: span 0-63 level CPU
groups: 0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60 (cpu_power = xxxx) 1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61 (cpu_power = xxxx) 2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62 (cpu_power = xxxx) 3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63 (cpu_power = xxxx)
not sure what the best way to present this is but 64 copies of this output do consume a lot of log buffer space!
* misc stuff, where we get 64 copies of a single message, eg
Switched to high resolution mode on CPU 0
ACPI: Processor [CPU0] (supports 8 throttling states) processor LNXCPU:01: registered as cooling_device1
Commit-ID: e23a8b6a8f319c0f08b6ccef2dccbb37e7603dc2 Gitweb: http://git.kernel.org/tip/e23a8b6a8f319c0f08b6ccef2dccbb37e7603dc2 Author: Roland Dreier <[email protected]> AuthorDate: Wed, 23 Sep 2009 15:35:35 -0700 Committer: Ingo Molnar <[email protected]> CommitDate: Thu, 24 Sep 2009 11:35:19 +0200
x86: Reduce verbosity of "PAT enabled" kernel message
On modern systems, the kernel prints the message
x86 PAT enabled: cpu 0, old 0x7040600070406, new 0x7010600070106
once for every CPU.
This gets kind of ridiculous on huge systems; for example, on a 64-thread system I was lucky enough to get:
dmesg| grep 'PAT enabled' | wc
64 704 5174
There is already a BUG() if non-boot CPUs have PAT capabilities that don't match the boot CPU, so just print the message on the boot CPU. (I kept the print after the wrmsrl() that enables PAT, so that the log output continues to mean that the system survived enabling PAT on the boot CPU)
LKML-Reference: <[email protected]>
--- arch/x86/mm/pat.c | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-)
arch/x86/mm/pat.c +5 -2
@@ -81,6 +81,7 @@ enum {void pat_init (void ) {u64 pat; +bool boot_cpu = !boot_pat_state;if (!pat_enabled)return ;
@@ -122,8 +123,10 @@ void pat_init(void)rdmsrl (MSR_IA32_CR_PAT , boot_pat_state);wrmsrl (MSR_IA32_CR_PAT , pat); -printk (KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n" , -smp_processor_id (), boot_pat_state, pat); + +if (boot_cpu) +printk (KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n" , +smp_processor_id (), boot_pat_state, pat); }#undef PAT
Quoted 2 lines · Roland Dreier
By the way, there is further work to be done in this direction... these were the easy ones (that save me 127 lines of dmesg :).
thanks Roland - i've applied them to tip:x86/urgent, for v2.6.32 merge.
Feel free to send more such patches, they are nice improvements.
Quoted 21 lines
The philosophy should probably be things that we would want to see to debug crashes at boot should go into the kernel log -- that makes it easy for users to grab the output and send it on for debugging. If we can move other stuff somewhere that users can grab easily after boot, then that's probably worth it.
I think the next things to look at are:
* secondary CPU booting:
Calibrating delay using timer specific routine.. xxxx BogoMIPS (lpj=xxxx)
CPU: Physical Processor ID: 2
CPU: Processor Core ID: 0
CPU: L1 I cache: xxxx, L1 D cache: xxxx
CPU: L2 cache: xxxx
CPU: L3 cache: xxxx
CPU 1/0x40 -> Node 0
mce: CPU supports xxxx MCE banks
CPU1: Thermal monitoring enabled (TM1)
CPU 1 MCA banks ....
Quoted 4 lines
where all 64 CPUs are exactly the same -- I guess checking for non-homogenous systems might be kind of complicated but we really shouldn't print all this for every CPU when they all match the boot CPU precisely.
Agreed.
Quoted 11 lines
* scheduler domain stuff:
CPU0 attaching sched-domain:
domain 0: span 0,32 level SIBLING
groups: 0 (cpu_power= xxxx) 32 (cpu_power= xxxx)
domain 1: span 0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60 level MC
groups: 0,32 (cpu_power= xxxx) 4,36 (cpu_power= xxxx) 8,40 (cpu_power= xxxx) 12,44 (cpu_power= xxxx) 16,48 (cpu_power= xxxx) 20,52 (cpu_power= xxxx) 24,56 (cpu_power= xxxx) 28,60 (cpu_power= xxxx)
domain 2: span 0-63 level CPU
groups: 0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60 (cpu_power= xxxx) 1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61 (cpu_power= xxxx) 2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62 (cpu_power= xxxx) 3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63 (cpu_power= xxxx)
Quoted 2 lines
not sure what the best way to present this is but 64 copies of this output do consume a lot of log buffer space!
Just turn it off after the first instance and give it a sched=debug boot parameter perhaps - that's enough.
Quoted 6 lines
* misc stuff, where we get 64 copies of a single message, eg
Switched to high resolution mode on CPU 0
ACPI: Processor [CPU0] (supports 8 throttling states) processor LNXCPU:01: registered as
cooling_device1
Yeah. The switched-to message could be eliminated altogether.
Thanks,
Ingo
PATCH 1/1 hrtimer: Remove overly verbose "switch to high res mode" message
On big systems, printing <number of CPUs> copies of
Switched to high resolution mode on CPU nnn
clutters up the kernel log for minimal gain. Just get rid of them.
---
Quoted 1 line
Yeah. The switched-to message could be eliminated altogether.
easy enough...
kernel/hrtimer.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-)
kernel/hrtimer.c +0 -2
@@ -720,8 +720,6 @@ static int hrtimer_switch_to_hres(void)/* "Retrigger" the interrupt to get things going */ retrigger_next_event (NULL );local_irq_restore (flags); -printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());return 1 ; }
Re: [PATCH] hrtimer: Remove overly verbose "switch to high res mode" message
Quoted 9 lines · Roland Dreier
On big systems, printing <number of CPUs> copies of
Switched to high resolution mode on CPU nnn
clutters up the kernel log for minimal gain. Just get rid of them.Signed-off-by Roland Dreier <[email protected]>kernel/hrtimer.c +0
-0
kernel/hrtimer.c static int hrtimer_switch_to_hres(void) · line 724
Quoted 5 lines
kernel/hrtimer.c +0
-2@@ -720,8 +720,6 @@ static int hrtimer_switch_to_hres(void)/* "Retrigger" the interrupt to get things going */ retrigger_next_event (NULL );local_irq_restore (flags); -printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());
Perhaps pr_debug instead?
It gets compiled to nothing unless DEBUG is defined.
Re: [PATCH] hrtimer: Remove overly verbose "switch to high res mode" message
Quoted 6 lines
kernel/hrtimer.c static int hrtimer_switch_to_hres(void) · line 724
Quoted 3 lines
kernel/hrtimer.c +0
-2-printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());Perhaps
pr_debuginstead?
It gets compiled to nothing unless DEBUG is defined.
Dunno... is that kernel log message ever useful now?
- R.
Re: [PATCH] hrtimer: Remove overly verbose "switch to high res mode" message
Quoted 4 lines · Roland Dreier
Quoted 3 lines
kernel/hrtimer.c static int hrtimer_switch_to_hres(void) · line 724
Quoted 2 lines
kernel/hrtimer.c +0
-2-printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());Perhaps
pr_debuginstead?Dunno... is that kernel log message ever useful now?
Dunno either.
I think lots of kernel log messages are not particularly useful. Others frequently think different...
cheers, Joe
Re: [PATCH] hrtimer: Remove overly verbose "switch to high res mode" message
In reply to Roland DreierQuoted 8 lines · Roland Dreier
Quoted 6 lines
kernel/hrtimer.c static int hrtimer_switch_to_hres(void) · line 724
Quoted 3 lines
kernel/hrtimer.c +0
-2-printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());Perhaps
pr_debuginstead?
It gets compiled to nothing unless DEBUG is defined.Dunno... is that kernel log message ever useful now?
There's 3 cases to consider:
1) Kernel built with CONFIG_HIGH_RES_TIMERS=n. We get no message then.
2) We have high-res timers and it succeeded. We probably don't care if that's the usual case, so lose the message in that case, just like we probably don't need a message that we managed to fork off init.
3) We attempted to switch to high-res mode, and it *failed*. We already throw the appropriate KERN_WARNING for that a few lines up.
And since we issue a message in the one case we *care* about, we can determine that the *lack* of that KERN_WARNING means the switch went OK.
Thus, we don't need the message, as it doesn't *tell* us anything we didn't already know.
[tip:timers/urgent] hrtimer: Remove overly verbose "switch to high res mode" message
In reply to Roland DreierCommit-ID: d3f6302e7e51b41af86c6496ffb2f95e8f2179df Gitweb: http://git.kernel.org/tip/d3f6302e7e51b41af86c6496ffb2f95e8f2179df Author: Roland Dreier <[email protected]> AuthorDate: Thu, 24 Sep 2009 14:07:55 -0700 Committer: Ingo Molnar <[email protected]> CommitDate: Sat, 26 Sep 2009 16:58:09 +0200
hrtimer: Remove overly verbose "switch to high res mode" message
On big systems, printing <number of CPUs> copies of
Switched to high resolution mode on CPU nnn
clutters up the kernel log for minimal gain. Just get rid of them.
LKML-Reference: <[email protected]>
--- kernel/hrtimer.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-)
kernel/hrtimer.c +0 -2
@@ -720,8 +720,6 @@ static int hrtimer_switch_to_hres(void)/* "Retrigger" the interrupt to get things going */ retrigger_next_event (NULL );local_irq_restore (flags); -printk (KERN_DEBUG "Switched to high resolution mode on CPU %d\n" , -smp_processor_id ());return 1 ; }