On 15/06/07, Jürgen Keil <[EMAIL PROTECTED]> wrote:
> > Nevada Build 65 performace vs. Debian 4.0 performace is abysmal.
> >
> > - boot up process is 30 seconds with Debian 4.0 / ~ 50 with Nevada
>
> Yep, and it seems the 64-bit S-x86 kernel is worse than the 32-bit kernel,
> at least on multiprocessor (e.g. dual core) machines.
>
> The 64-bit S-x86 kernel appears to spend (waste?) 5-7 seconds inside
> release_bootstrap(), depending on cpu speed:
Joe B. has filed this issue as bug 6569219
"wasted time at amd64 boot - issues with htable_walk()"
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6569219
The fix he proposed (see bug #6569219) does indeed speed up
the amd64 boot process by 5-7 seconds.
Before applying the fix for #6569219, it had this in /var/adm/messages
(note the 5 seconds jump in time after the kb8042 driver message
and before the next pcplusmp message):
Jun 12 10:26:07 tiger2 genunix: [ID 936769 kern.notice] kb80420 is /isa/[EMAIL
PROTECTED],60/[EMAIL PROTECTED]
Jun 12 10:26:12 tiger2 pcplusmp: [ID 803547 kern.info] pcplusmp:
pciclass,0c0320 (ehci) instance 0 vector 0x15 ioapic 0x2 intin 0x15 is bound to
cpu 0
Jun 12 10:26:17 tiger2 npe: [ID 236367 kern.notice] PCI Express-device:
pci1043,[EMAIL PROTECTED],1, ehci0
Jun 12 10:26:17 tiger2 genunix: [ID 936769 kern.notice] ehci0 is /[EMAIL
PROTECTED],0/pci1043,[EMAIL PROTECTED],1
Jun 12 10:26:22 tiger2 usba: [ID 912658 kern.notice] USB 2.0 device
(usbd49,3210) operating at hi speed (USB 2.x) on USB 2.0 root hub: [EMAIL
PROTECTED], scsa2usb0 at bus address 2
Jun 12 10:26:22 tiger2 usba: [ID 349649 kern.notice] Maxtor Corporation 3200
604010123768
Jun 13 10:30:31 tiger2 i8042: [ID 526150 kern.notice] 8042 device: [EMAIL
PROTECTED], kb8042 # 0
Jun 13 10:30:31 tiger2 genunix: [ID 936769 kern.notice] kb80420 is /isa/[EMAIL
PROTECTED],60/[EMAIL PROTECTED]
Jun 13 10:30:36 tiger2 pcplusmp: [ID 803547 kern.info] pcplusmp:
pciclass,0c0320 (ehci) instance 0 vector 0x15 ioapic 0x2 intin 0x15 is bound to
cpu 0
Jun 13 10:30:41 tiger2 npe: [ID 236367 kern.notice] PCI Express-device:
pci1043,[EMAIL PROTECTED],1, ehci0
Jun 13 10:30:41 tiger2 genunix: [ID 936769 kern.notice] ehci0 is /[EMAIL
PROTECTED],0/pci1043,[EMAIL PROTECTED],1
Jun 13 10:30:46 tiger2 usba: [ID 912658 kern.notice] USB 2.0 device
(usbd49,3210) operating at hi speed (USB 2.x) on USB 2.0 root hub: [EMAIL
PROTECTED], scsa2usb0 at bus address 2
Jun 13 10:30:46 tiger2 usba: [ID 349649 kern.notice] Maxtor Corporation 3200
604010120544
With the fix for #6569219 the pcplusmp message is now logged during
the same second as the kb8042 message:
Jun 15 10:36:33 tiger2 i8042: [ID 526150 kern.notice] 8042 device: [EMAIL
PROTECTED], kb8042 # 0
Jun 15 10:36:33 tiger2 genunix: [ID 936769 kern.notice] kb80420 is /isa/[EMAIL
PROTECTED],60/[EMAIL PROTECTED]
Jun 15 10:36:33 tiger2 pcplusmp: [ID 803547 kern.info] pcplusmp:
pciclass,0c0320 (ehci) instance 0 vector 0x15 ioapic 0x2 intin 0x15 is bound to
cpu 0
Jun 15 10:36:38 tiger2 npe: [ID 236367 kern.notice] PCI Express-device:
pci1043,[EMAIL PROTECTED],1, ehci0
Jun 15 10:36:38 tiger2 genunix: [ID 936769 kern.notice] ehci0 is /[EMAIL
PROTECTED],0/pci1043,[EMAIL PROTECTED],1
Jun 15 10:36:43 tiger2 usba: [ID 912658 kern.notice] USB 2.0 device
(usbd49,3210) operating at hi speed (USB 2.x) on USB 2.0 root hub: [EMAIL
PROTECTED], scsa2usb0 at bus address 2
Jun 15 10:36:43 tiger2 usba: [ID 349649 kern.notice] Maxtor Corporation 3200
604010120546
Now I'm going to have a look at the ehci USB2.0 driver; I seems that it needs
5 seconds to "initialize" (time spent between the "npe: .. ehci0" message
and "genunix: ehci0 is ...").
And the USB disk that happens to be connected to that ehci USB 2.0 controller
delays the boot by another five seconds...
Which makes me wonder, is there anything we can do to "background"
certain device detection probing so that the login process can start?
In other words, do we really need to delay the boot process to detect
USB storage devices unless one of the devices is required to be
mounted *at boot* or is the boot device itself?
--
"Less is only more where more is no good." --Frank Lloyd Wright
Shawn Walker, Software and Systems Analyst
[EMAIL PROTECTED] - http://binarycrusader.blogspot.com/
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
