wary511_update-1, plus more

I reported just before releasing Wary 5.1.1.53 that I had to revert insertion of rerwin's wary511_update-1.pet in Woof, as it broke the sound on my laptop. I have now found out why the sound was broken and have fixed it, and completed integration of rerwin's PET into Woof.

The original blog report on wary511_update-1:
http://bkhome.org/archive/blog2/201105/rerwin-wary511update-1.html

Blog reports on partial re-implementation:
http://bkhome.org/archive/blog2/201105/fixes-for-pupeventbackendmodprobe.html
http://bkhome.org/archive/blog2/201105/fixes-for-etcmodprobedalsaconf.html

By implementing the PET step by step I was able to find out what broke my sound. It was the agrsm.tar.gz firmware tarball, which has a new "sleep 2" in the installation script. This delayed the creation of /etc/modprobe.d/modem_agrsm.conf which has some blacklist entries for sound modules. Consequently it appears that one or more of these sound modules was loading, which upset the sound. Note though, this problem was only on first boot and would have gone away on subsequent boots.

I have now fully integrated rerwin's PET into Woof.

Extreme timing bug
See point 3 in rerwin's second post on this forum page:
http://murga-linux.com/puppy/viewtopic.php?t=65028&start=150

Rerwin has a DGC USB modem for which the kernel generates several "add" uevents for loading the same module, in rapid succession, and this is an extreme stress test. My /sbin/pup_event_backend_modprobe was not quite up to it, and failed to load the dgcusbdcp.ko module.

I have implemented a fix for this, in /sbin/pup_event_backend_modprobe. Rerwin sent a pm that he will get back to me soon, after trying it.

---

Comments

Kernel build script
Username: Iguleder
Barry - 2.6.32.40 was released today and has the DMA fix from Arch integrated, among other things. I'm working on a generic build script for 2.6.32.x kernels at the moment - it downloads the latest kernel sources, the latest Aufs for it, patches Aufs, patches the kernel and produces a directory with all sources and both the headers and actual kernel packages. It has a nice feature that patches the kernel (using sed) to make it appear as 2.6.32, so the modem drivers, etc' can be compiled just once, for the first kernel built this way. Could be very useful for Wary, because of all the modem drivers and the LTS nature of its kernel. I'll upload it to the forum once it's full working.

Re kernel build
Username: BarryK
"Iguleder, It will be very nice to have it automated!!! It is good, the guy who is maintaining 2.6.32.x is keeping up his enthusiasm and updating regularly.

Re kernel build
Username: BarryK
"There is one downside to upgrading the kernel, we also have to recompile the nVidia (and maybe ATI) video modules. If that could be automated also, then heaven!

Nope!
Username: Iguleder
"Nope - if the driver is built against a kernel built this way, it will be present in lib/modules/2.6.32, which means it should work against any 2.6.32.x built using the script. It's kind of impossible to automate the nVidia installer, because you need to reboot and use a save file to catch the files it installs, but there's no need to do that if all 2.6.32.x kernels are "compatible" :) I'm building a Lupu 5.2.5 with 2.6.39.40 at the moment. The script looks good, added a build log and the patching works well.

It's ready
Username: Iguleder
"The script is ready, you can find it [url=http://murga-linux.com/puppy/viewtopic.php?p=521472]here[/url] and a the whole output [url=http://iguleder.info/puppy/kernel/linux_kernel-2.6.32-40-lts/dist]here[/url], except kernel_sources-2.6.32-40-lts.sfs, which is too big for me to upload. Writing this from Conservative Lucid Puppy 5.2.5 (lupc-525.iso) :)

Module loading fixed
Username: BarryK
"Rerwin has confirmed that my fix in /sbin/pup_event_backend_modprobe has fixed the problem he had of a DGC modem module not loading. More very good news.

Tags: woof