I guess the rawhide kernel has extra checks/debugging turned on - which can slow things down.
Any idea what these are?
I'd like to disable these patches and rebuild the kernel [rawhide kernels currently being more stable on my T40]
I'd appreciate any pointers.
Thanks, Satish
On Fri, 18 Feb 2005 12:27:36 -0600 (CST), Satish Balay balay@fastmail.fm wrote:
I guess the rawhide kernel has extra checks/debugging turned on - which can slow things down.
Any idea what these are?
Taking a quick gander through my /boot/config-2.6.10-1.1143_FC4...
# # Kernel hacking # CONFIG_DEBUG_KERNEL=y CONFIG_MAGIC_SYSRQ=y # CONFIG_SCHEDSTATS is not set CONFIG_DEBUG_SLAB=y CONFIG_DEBUG_SPINLOCK=y CONFIG_DEBUG_SPINLOCK_SLEEP=y # CONFIG_DEBUG_KOBJECT is not set CONFIG_DEBUG_HIGHMEM=y CONFIG_DEBUG_BUGVERBOSE=y CONFIG_DEBUG_INFO=y # CONFIG_DEBUG_FS is not set # CONFIG_FRAME_POINTER is not set CONFIG_EARLY_PRINTK=y CONFIG_DEBUG_STACKOVERFLOW=y CONFIG_KPROBES=y CONFIG_DEBUG_STACK_USAGE=y CONFIG_DEBUG_PAGEALLOC=y CONFIG_4KSTACKS=y
Probably just set all the CONFIG_DEBUG*=y there to =n. I'm not sure which would affect speed the most, but I'd guess disabling CONFIG_DEBUG_SLAB will cut down most of the extraneous memory usage (if that's as noticeable for you as it is for me on my 256 MB machine).
The more debugs you keep on, though, the more chance you have of being able to usefully report any problems...
(cc:ing you because I'm not sure I have my subscription set up to allow posting from this address)
--wes
On Fri, 18 Feb 2005, Wes Shull wrote:
Taking a quick gander through my /boot/config-2.6.10-1.1143_FC4...
Thanks for the pointer. I was earlier looking for difference in patches in the spec file.
Looking at the diffs in config with 766_FC3 [latest FC3 update] & 1143 - I see the following:
CONFIG_DEBUG_SLAB=y CONFIG_DEBUG_BUGVERBOSE=y CONFIG_DEBUG_PAGEALLOC=y
I'll disable these - and see if it makes a difference. The others in your list are:
CONFIG_DEBUG_KERNEL=y CONFIG_DEBUG_HIGHMEM=y CONFIG_DEBUG_SPINLOCK=y CONFIG_DEBUG_SPINLOCK_SLEEP=y CONFIG_DEBUG_INFO=y CONFIG_DEBUG_STACKOVERFLOW=y CONFIG_DEBUG_STACK_USAGE=y CONFIG_DEBUG_PAGEALLOC=y
CONFIG_MAGIC_SYSRQ=y CONFIG_EARLY_PRINTK=y CONFIG_KPROBES=y CONFIG_4KSTACKS=y
I guess I should try disabling some of these as well.
Satish
On Fri, Feb 18, 2005 at 02:22:27PM -0600, Satish Balay wrote:
Thanks for the pointer. I was earlier looking for difference in patches in the spec file.
Looking at the diffs in config with 766_FC3 [latest FC3 update] & 1143
- I see the following:
CONFIG_DEBUG_SLAB=y CONFIG_DEBUG_BUGVERBOSE=y CONFIG_DEBUG_PAGEALLOC=y
Uhhh... Aren't CONFIG_DEBUG_SLAB and CONFIG_DEBUG_PAGEALLOC incompatible with each other? At least, that's what I was told by none other than Dave Jones himself (RH Bugzilla #144324, comment #15).
That reminds me, I have one FC3 system here running rawhide kernels, and it keeps oopsing once every couple of days. Unfortunately I haven't had a chance to set up a serial console on it again (I'll do that later today), but I wonder if the combination of DEBUG_SLAB and DEBUG_PAGEALLOC is why it's oopsing...
-Barry K. Nathan barryn@pobox.com
On Fri, 18 Feb 2005, Barry K. Nathan wrote:
Uhhh... Aren't CONFIG_DEBUG_SLAB and CONFIG_DEBUG_PAGEALLOC incompatible with each other? At least, that's what I was told by none other than Dave Jones himself (RH Bugzilla #144324, comment #15).
Chagelog has the following entry:
* Sun Jan 23 2005 Dave Jones davej@redhat.com
- Updated periodic slab debug check from Manfred. - Enable PAGE_ALLOC debugging again, it should now be fixed. - 2.6.11-rc2-bk1
Satish
On Fri, Feb 18, 2005 at 03:47:27PM -0600, Satish Balay wrote:
Chagelog has the following entry:
- Sun Jan 23 2005 Dave Jones davej@redhat.com
- Updated periodic slab debug check from Manfred.
- Enable PAGE_ALLOC debugging again, it should now be fixed.
- 2.6.11-rc2-bk1
Doh! Thanks for setting me straight. (I was sleep-deprived and in a hurry, so I forgot to check the changelog.)
-Barry K. Nathan barryn@pobox.com