Woof uploaded, July 12, 2011

This is commit '20110712172303'.
Previous upload was on July 4:
http://bkhome.org/blog/?viewDetailed=02331

This Woof has modified /sbin/wait4usb and /init scripts in the initramfs, and modified /etc/rc.d/rc.sysinit in the main filesystem, to support the 2.6.39-3 kernel with hid and usb core drivers (including usb-storage) builtin to the kernel.

The scripts also support the kernel without my usb-storage patch. That patch was used in 2.6.34+ kernels, see 'usb-storage.patch.gz' at:
http://bkhome.org/sources/kernel-2.6.35.7/patches/

Without that patch, we don't have any means to determine when all usb drives have been fully scanned and are ready for use. Instead, I have put a "sleep 5" into /sbin/wait4usb. This is very crude, but the kernel developers took away the only means that I know of to know when all usb drives are ready, with 2.6.34+ -- this issue has been reported in this blog.

For experimentation, wait4usb writes to /initrd/tmp/usb_drvs_found_1sec_intervals each second, so we can get a report on how long it takes before drives appear in /sys/block.

Booting from CD on my Pentium4 desktop, with a usb pen drive plugged in, /initrd/tmp/usb_drvs_found_1sec_intervals has this:

USBPCI: 0C0320 0C0310 0C0310 0C0300 0C0300 

USBDRVS:

USBPCI: 0C0320 0C0310 0C0310 0C0300 0C0300
USBDRVS:

USBPCI: 0C0320 0C0310 0C0310 0C0300 0C0300
USBDRVS: sdc

USBPCI: 0C0320 0C0310 0C0310 0C0300 0C0300
USBDRVS: sdc

USBPCI: 0C0320 0C0310 0C0310 0C0300 0C0300
USBDRVS: sdc


...which means that sdc appears after 2 seconds.

This might be fine for modern hardware, but what about USB1 interfaces? What about USB optical drives? -- unfortunately mine was stolen so can't test that.

Tonight I am going to recompile the 2.6.39.3 kernel, with a few small changes. I am also going to test with my "usb-storage" patch put back in.

Note, I have built Wary 5.1.2.3 with the 2.6.39-3 kernel, uploading it now. If anyone is interested in trying it, to see if it detects their usb ok, you're welcome. Posting that /initrd/tmp/usb_drvs_found_1sec_intervals would be most helpful. I'll post when it is uploaded.


Posted on 12 Jul 2011, 18:02


Comments:

Posted on 12 Jul 2011, 18:38 by 01micko
re woof usb
Woohoo! Thanks Barry.

Grabbing woof now and gonna try my 2.6.37.6 WITH your usb_storage patch applied and the Linus 1 second patch backed out (reverted to 5 seconds).

Got to get up early tomorrow so if possible will post results tonight, if not tomorrow afternoon.
Cheers!




Posted on 12 Jul 2011, 19:31 by pemasu
2.6.39-3 naming
Will you include that new 2.6.39-3 naming syntax to the /lib/modules/kernel naming with symlink ?

I downloaded already the latest woof also.


Posted on 12 Jul 2011, 20:30 by BarryK
2.6.39 naming
pemasu,
The directory is /lib/modules/2.6.39, regardless what sub-version. This is how we are doing it, thanks to Iguleder who paved the way. Ditto for the 2.6.32 series. No symlink is required.





Posted on 12 Jul 2011, 20:34 by BarryK
Wary 5.1.2.3
For testing, Wary built with latest Woof and 2.6.39-3 kernel (111.9MB):

http://distro.ibiblio.org/quirky/test/wary-5.1.2.3.iso

Md5sum: e6ca204e10c080092fb68d153a446ce8

This does not represent the next release of Wary, just a build for testing USB detection, for anyone who is interested.



Posted on 12 Jul 2011, 20:39 by 01micko
it works
So I woofed up an spup in record time, actually I should upload it for further testing , maybe tomorrow.

At the time of the woof announce I was compiling QT.. I didn't want to interrupt that, new2dir is still running this minute, so I did a "replacement" of a former USB install which was known NOT to work on this troublesome compaq machine.. needless to say I am posting from the troublemaker now! (more tests tomorrow, and note, the same 2.6.37.6 I compiled Sunday to which I have been referring)
Thanks Barry!!!


Posted on 12 Jul 2011, 23:44 by pemasu
recompiling 2.6.39 kernel
There seems to be unsatisfied dependency timberdale in your compiled kernel. When the recompiling is finished I track it and try to fix it. I will use intel atom optimization. 586 optimization seems to cause quite excessive cpu activity with low usage.


Posted on 13 Jul 2011, 2:10 by pakt
re: usb_drvs_found_1sec_intervals
Booting from CD on my oldest laptop, a 300MHz PII with USB1.1, with a USB pen drive plugged in:

USBPCI: 0C0300
USBDRVS: sdb

USBPCI: 0C0300
USBDRVS: sdb

USBPCI: 0C0300
USBDRVS: sdb

USBPCI: 0C0300
USBDRVS: sdb

USBPCI: 0C0300
USBDRVS: sdb



Posted on 13 Jul 2011, 3:00 by Sage
W5.1.2.3beta-usb
Everything working and fast. USB 1.1 & 2.0 devices sign on within 3s. D/l-ed and installed Opera 11.5 as 'otherDeb' and had it running within 30s, even on a slow secondary hub. Best yet.


Posted on 13 Jul 2011, 4:03 by GCMartin
Testing the WARY Alpha
Barry, I've just completed testing of your WARY on both a 384MB old P3 and a 4GB X2. Its happy and operational on each.

I know this is not the place to post WARY reports, so none are included.

USB
I have a USB stick that has a data and a Swap partition on it. Both WARY boots finds it. I have never save-session to HDD/USB, but Live media (CD/DVD0 save-session is working in this arrangement.

Hope it helps.
P.S. Check PM



Posted on 13 Jul 2011, 5:40 by BarryK
re feedback
Guys, thanks for the feedback.

I have recompiled the kernel with my usb-storage patch, and that works too.

I have modified 'wait4usb' to work with or without the usb-storage patch. I set it so that the fastest it can run through is 3 seconds, worst case 7 seconds, so bootup will in most cases be faster.

pemasu,
Note that your 2.6.38.x config file has CONFIG_PM_DEBUG and CONFIG_PM_VERBOSE set, which makes the system log too verbose. I used your .config as the basis for 2.6.39-3, but in latest compile I have disabled those two.

Note also, you have "Enhanced RTC" disabled. This is something that I am uncertain about, anyway I enabled it.

Yeah, I noticed that Timberdale problem.



Posted on 13 Jul 2011, 5:57 by james c
wary 5.1.2.3

Wary 5.1.2.3 w/ 2.6.39.3 seems to be working good here as well.


Posted on 13 Jul 2011, 8:36 by pemasu
2.6.39 kernel feedback
About those debug enabled features. I enabled them because I thought that Powertop application might benefit of them so that it gets more accurate information. But now I know that it is debug kernel which is needed for Powertop. CONFIG_DEBUG_KERNEL=y and CONFIG_TIMER_STATS

Enhanced RTC disabled was not consciously done. Thank you of remind. I will enable that in the future.

I fixed the missing dependencies with my intel atom optimization version.

Thank you of feedback. Highly appreciated. And thank you that you have used precious time with boot time usb detection. It has given me some gray hairs.


Posted on 13 Jul 2011, 8:53 by BarryK
486/586 kernel
pemasu,
I have always configured for i486 CPU, because that is the minimum that I intended Puppy to run on. However, this time I decided the minimum is a i586.



Posted on 13 Jul 2011, 11:15 by steev
PAE and Pentium M
With PAE enabled, this kernel may not boot on some versions of the Pentium M (i686?)

Some Slackware users had problems when PAE was first enabled in the "smp" kernels:
http://www.linuxquestions.org/questions/slackware-14/cannot-boot-current-have-done-something-stupid-882943/



Posted on 13 Jul 2011, 15:23 by BarryK
pae seems ok
steev,
Yes, that is interesting. I wonder if my choice of i586 CPU and PAE has something to do with success for us, or maybe the 2.6.39.x kernel has improved PAE compatibility.

I have tested it on a Pentium4 and Celeron CPU, works, and above reports show success back to a PII.

I wonder about the Edubook, which has an RDC 586-class CPU? I'll try that.



Posted on 13 Jul 2011, 16:16 by pemasu
PAE and noncompatibility
There has been Medion 8188 with dual core processor which has needed nolapic before PAE enabled kernel Puppy has been booted. That boot time parameter, nolapic usage, has had unwanted consequences like only one processor used.
I dont know whose dual core processor it is.

PAE adds about 18-20 Mb of ram usage in several reports and that memory usage addition has been the same with several cpu optimizations like i486, intel atom, dual core and up.



Posted on 13 Jul 2011, 16:34 by BarryK
PAE does not work on Edubook
Ok, now it is becoming clearer. The PAE kernel does not work on the Edubook, which has an "i586" CPU.
Quoting from the link posted by steev:

PAE is provided by Intel Pentium Pro and above CPUs, including all later Pentium-series processors (except the 400 MHz-bus versions of the Pentium M). It is also available on other processors with similar or more advanced versions of the same architecture, such as the AMD Athlon[dubious discuss] and later AMD processor models.

In that case, I am going to recompile the kernel again tonight, this time specify CPU architecture as Pentium Pro, that being the minimum that supports PAE.
Although, again from that link, Slackware kernel has this when updating to PAE:

HIGHMEM4G y -> n
HIGHMEM64G n -> y
LOG_BUF_SHIFT 15 -> 18
M686 y -> n
MPENTIUMIII n -> y





Posted on 13 Jul 2011, 17:04 by gjuhasz
No sound
Runs well on P-II Toshiba Tecra-8000 machine
except that the Yamaha internal sound card (opl3sa2) is not recognized


Posted on 13 Jul 2011, 17:10 by gjuhasz
No sound
Wary 5.1.2.3 w/ 2.6.39.3 kernel runs well on my P-II Toshiba Tecra-8000 machine
except that the Yamaha internal sound card (opl3sa2) is not recognized at all.
Compared to Wary 5.1.2, driver(s) missing.



Posted on 13 Jul 2011, 17:24 by pemasu
PAE and cpu choice
Oops. I didnt mean to i486. It is not possible. When you choose cpu, the PAE possibility is exluded if it is not possible. Take i486 and you dont get possibility to choose PAE.


Posted on 13 Jul 2011, 21:36 by gjuhasz
ROX right-click problem
I see that the old ROX right-click menu problem appears again, not only in Wary 5.1.2.3 but in all puppies based on the latest woof.


Posted on 17 Jul 2011, 10:05 by 01micko
Browser Choice
Hi Barry
I notice in this woof there is no choice of whether to have a zdrv or not. That's fine but I have an idea. I guess you are familiar with technosaurus' 'zdrv cutter'. What if the there was an option in woof to include the browser in the zdrv? One of the biggest complaints about Puppy is that if you update the browser then the old browser is just dead weight in the main sfs. I think this would be the ideal solution to that scenario. After you install a new browser, either as an sfs or a .pet then you could use zdrv-cutter to weed out the dead weight.
Thoughts?



Posted on 22 Jul 2011, 12:23 by archwndas
Sandy bridge 2 and hybrid graphics
Hi Barry,

Thanks for the new 39-3 kernel version. I am wondering however if you took into account the latest problems of the linux kernel with hybrid graphics in the new Sandy bridge 2 based laptops. I have one such and kernel 39-2 crashes from time to time after some reboot. Both graphics cards are on and although you can switch the radeon off by:

echo OFF > /sys/kernel/debug/vgaswitcheroo/switch

this causes panics after 2 reboots, and you need to press power off and reboot and again power off and reboot until the kernel stops crashing and you finally get back to the gnome login screen. This is the case with all the 2.38-10 based Ubuntu distros and Fedora and Debian Sid 2.39.2 I tried so far. Neither FreeBSD nor Solaris the latest can boot to X from their live CDs.

I know you can do magic. Have you any work-arounds?