You are not logged in.
Sorry for sidetracking the discussion - good work you do by the way! - but I would like some input on the following:
- have any of you tried an standard install to a good performance USB stick?
I wonder because my wife is mainly using this Aspire One, and she might not be too fond of sudden changes, but I would prefer Arch when I use it. Dual boot is... eh... yes not an option unless I change the hard-drive to a standard 1.8" ZIF one.
Since the performance of the SSD isn't something to sing arias about, I start to think that a dedicated USB stick would perform at least equally. The stick I have in mind is pretty cheap (at least here) but of good quality, Patriot Xporter 8GB. Even with larger files this little piece can keep a write speed of close to 20 MB/s and read speed over 30 MB/s, so I figured it might even be a faster option. I see no problem in having one USB stick solely for one purpose, considering how cheap these sticks are nowadays.
Any thoughts? Is it a stupid idea?
Offline
Hi all...
(I'm New in this forum)
Today I've add 1 GB to my Acer One but i can see only 1 GB instead of 1.5 GB in:
cat /proc/meminfo
MemTotal: 904400 kB
...
I'm trying both the gothicknight and carlocci kernels.
(The HIGHMEM4G is disabled)
[ 0.000000] Warning only 896MB will be used.
[ 0.000000] Use a HIGHMEM enabled kernel.
[ 0.000000] 896MB LOWMEM available.
Can I suggest to change the .config from:
CONFIG_NOHIGHMEM=y
# CONFIG_HIGHMEM4G is not set
# CONFIG_HIGHMEM64G is not set
to
# CONFIG_NOHIGHMEM is not set
CONFIG_HIGHMEM4G=y
# CONFIG_HIGHMEM64G is not set
CONFIG_HIGHMEM=y
CONFIG_ARCH_ENABLE_MEMORY_HOTPLUG=y
Thanks...
Last edited by CeccoPierangioliEugenio (2008-09-16 09:24:34)
Offline
This has no impact on "normal" systems. I'll upload the package in a few minutes.
UPDATE:
New build with support for up to 4GB of memory.
Deadline is now the default I/O scheduler (I think it works better, but some metrics would be great if someone can supply)
Kernel debug is disabled until related panics are solved.
Disabled ath5k, it doesn't work either way so, it's pointless to have it.
URL: http://clientes.netvisao.pt/gtknight/ke … pkg.tar.gz
Last edited by gothicknight (2008-09-16 13:00:04)
Offline
Great, everything is working now - even after suspend I installed Openbox with xcompmgr and it looks just awesome (and everything's smooth). I switched to deadline scheduler too and it removed the "hard locks", it still slows down every once in a while (with noop the hardisk led was constantly on when "locking", now it flashes).
Offline
The latest version works fine for me, too. Many thanks for taking the time to compile the kernel again and again. When I have the time, I will run some benchmark on battery life compared to the stock kernel. I do a lot of terminal work with lots of idle time. I think that the slack kernel can make a difference here, but 20% as reported above would be quite siginificant.
Now that everything is working fine, why not step up and live on the bleeding edge with a 2.6.27-rc6 kernel? I read somewhere that ath5k should work in 2.6.27 at an unstable/experimental level, we could try that.
Offline
What about support for countrycode in ath_pci module? Loading it with countrycode value other than 0 make wlan not working...
I need support for above than 11 channels
thx in advance
kTT
Offline
Humm... 2.6.27 trully looks nice to have.
You have UBIFS which has been in my mind to use since I've first start searching for flash filesystems. I think a develop branch can be in use. I'll try to create tomorow (If everything is calm enough) the PKGBUILD and test it a little.
Although I think that we have still to find the reason behing the kernel panics with kernel debug enabled
Offline
Although I think that we have still to find the reason behing the kernel panics with kernel debug enabled
I'd be glad to help, but I don't know how much I could be of use. I could transcribe the kernel panic message and lines preceding, but I wouldn't know what to do with the information other than to give it to someone else (who could probably get the same info themselves, if they have a One).
Offline
Humm... 2.6.27 trully looks nice to have.
You have UBIFS which has been in my mind to use since I've first start searching for flash filesystems. I think a develop branch can be in use. I'll try to create tomorow (If everything is calm enough) the PKGBUILD and test it a little.Although I think that we have still to find the reason behing the kernel panics with kernel debug enabled
Would UFIFS be any use? IIRC, the SSD in the AA1 only presents a block device interface, while UBIFS runs on top of UBI, which in turn requires MTD... which means that it would be useless on top of the PATA interface. However, according to the Intel Z-P230 datasheet, the SSD shipped with the Aspire One performs its own wear-levelling, and has an effective lifespan of three years based on a 60% random workload of 2GB/day.
Last edited by bremac (2008-09-16 22:11:09)
Offline
/facepalm
In a response to your questions gothicknight, I was confirming which kernel I was using, and it seems that I've been using the same kernel I installed Arch with all those months ago. I wasn't aware till some forum + wiki searching that you have to modify your grub setup to load a different kernel as this one is named differently than the original.
So, yeah, it's suspending fine now via echo "mem" > /sys/power/state and pm-suspend, from the commandline and X. I had the same issues as others listed here regarding sound, but the details on fixes list here remedied that also.
Thanks for your hard work on the kernels so far, and I thank you for the help you gave me (without even answering a question!)
Offline
I'd be glad to help, but I don't know how much I could be of use. I could transcribe the kernel panic message and lines preceding, but I wouldn't know what to do with the information other than to give it to someone else (who could probably get the same info themselves, if they have a One).
I'll try today to compile the new Kconfig with CONFIG_DEBUG_KERNEL enabled and see what's happening.
Would UFIFS be any use? IIRC, the SSD in the AA1 only presents a block device interface,
You're right, there's no need for such thing. Not even for the SDHC as they are also block devices.... My mistake
Thanks for your hard work on the kernels so far, and I thank you for the help you gave me (without even answering a question!)
Sorry for that, must have missed it...
UPDATE: I've tried to get the kernel panic but without success... No log of the panic is recorded, my screen isn't big enough to read it entirely and plugging a external monitor doesn't do the trick either
Last edited by gothicknight (2008-09-17 08:27:47)
Offline
Hello,
I'm currently an Ubuntu user, but I've been reading this thread as it has a lot of useful information. Keep up the good work! (and I might switch to Arch one day!)
What about support for countrycode in ath_pci module? Loading it with countrycode value other than 0 make wlan not working...
I need support for above than 11 channels
thx in advance
To get channels 1 - 13 when using wifi type:
modprobe ath_pci -r
modprobe ath_pci xchanmode=1or make new file named "madwifi" to /etc/modprobe.d/
with line:options ath_pci xchanmode=1
Other usable options for madwifi, type:
modinfo ath_pci
Warning: do not try overrun countrycode. it does not work!
My current install (gnome with 2.6.27) boots to the desktop in about 35 seconds. What are you experiences?
Offline
thx 2manydjs :)
my install boot takes about 23sec (grub->slim->openbox)
kTT
Offline
I've tried to get the kernel panic but without success... No log of the panic is recorded, my screen isn't big enough to read it entirely and plugging a external monitor doesn't do the trick either
Just had a thought. Is it possible that the kernel panic is related to using tmpfs for /var/log? That is a pretty nonstandard configuration, and I don't know, but it could be trying to write to a logfile before /var/log is in place. Anyways, I thought I remembered being able to scroll with shift+pgup/pgdn after the kernel panic. I'll sabotage my machine to see if that's true after work, and type out the info if it is.
V--Like I said, I don't know.
Last edited by qylvaran (2008-09-17 15:58:32)
Offline
gothicknight wrote:I've tried to get the kernel panic but without success... No log of the panic is recorded, my screen isn't big enough to read it entirely and plugging a external monitor doesn't do the trick either
Just had a thought. Is it possible that the kernel panic is related to using tmpfs for /var/log? That is a pretty nonstandard configuration, and I don't know, but it could be trying to write to a logfile before /var/log is in place. Anyways, I thought I remembered being able to scroll with shift+pgup/pgdn after the kernel panic. I'll sabotage my machine to see if that's true after work, and type out the info if it is.
I didn't think the kernel wrote any of its own log files - I thought that was syslog's job? In any case, writing before /var/log was in place would mean writing before / is in place, so I'm unsure how anything could even attempt such a thing. Of course, there could always be an (unlikely) bug in tmpfs itself which only emerges with a certain configuration and access pattern.
Offline
Sorry for that, must have missed it... hmm
It's all part of the process. Your request for additional information pointed me toward the pathway to find the answer. I'm still rather new, but learning fast
Offline
May I suggest to build snd-hda-intel as a module in the next version of the custom kernel? This way it would be possible to reload the module with alternating model=acer and model=auto to switch between working suspend and working internal microphone. The way it is now, the model option can only be given at boot time.
Btw, the gain in boot time is not that significant for me. udev is down to 3.6 sec, but in exchange, booting the kernel takes more time (~2sec). I think that is because the kernel is larger now with everything compiled in.
Offline
May I suggest to build snd-hda-intel as a module in the next version of the custom kernel? This way it would be possible to reload the module with alternating model=acer and model=auto to switch between working suspend and working internal microphone. The way it is now, the model option can only be given at boot time.
If it's a module anyway, why not just use the parameter model=auto and unload/reload it when suspending via /etc/pm/config.d/modules, instead of swapping back and forth? That way you get the best of both worlds.
Offline
If it's a module anyway, why not just use the parameter model=auto and unload/reload it when suspending via /etc/pm/config.d/modules, instead of swapping back and forth? That way you get the best of both worlds.
No, it's not a module in the lates kernel by gothicknight. But once it is again a module, your suggestion looks very interesting, I may try that. However, I remember having read somwhere that this reloading after suspend does not work with the alsa module and therefore model=acer has to be used.
Btw, I manage to get from Grub to X in 15 seconds with all daemons started in background and dwm as window manager.
Last edited by thomasd (2008-09-17 20:16:22)
Offline
No, it's not a module in the lates kernel by gothicknight. But once it is again a module, your suggestion looks very interesting
That's what I meant, though I can see how the chronology appears to be misrepresented.
However, I remember having read somwhere that this reloading after suspend does not work with the alsa module and therefore model=acer has to be used.
Hrm, I guess you're right... according to http://ubuntuforums.org/showthread.php?t=894541&page=2 loading/unloading snd_hda_intel doesn't work properly after multiple unload/reloads... in which case, I don't think having it as a module would really help. Then again, others in the same thread report that unloading/reloading the module works just fine on a different model of the EEEpc, so the problems may be more hardware- than alsa-related (if that's even possible...).
Offline
Hrm, I guess you're right... according to http://ubuntuforums.org/showthread.php?t=894541&page=2 loading/unloading snd_hda_intel doesn't work properly after multiple unload/reloads... in which case, I don't think having it as a module would really help. Then again, others in the same thread report that unloading/reloading the module works just fine on a different model of the EEEpc, so the problems may be more hardware- than alsa-related (if that's even possible...).
Sorry, I misread what you intended to say in your previous post. I will just try with the stock kernel whether it works with /etc/pm/config.d/modules. In the end, this will probably all be solved in a later kernel/alsa version.
EDIT: I cannot try it with the stock kernel, as the SD card will be corrupted after resume.
EDIT2: Just one more question to the people using the custom kernel. Do you use the same initrd line as the stock kernel in menu.lst? Or can that be omitted, because everything is compiled in kernel?
Last edited by thomasd (2008-09-18 07:05:24)
Offline
If I'm not mistaken you can change a parameter of a running module by use of sysctrl. But if that doesn't work, post a message and I can rebuild this.
Btw, the gain in boot time is not that significant for me. udev is down to 3.6 sec, but in exchange, booting the kernel takes more time (~2sec). I think that is because the kernel is larger now with everything compiled in.
For me I can see a gain in about ~6 seconds on udev, other than that it much the same to me.
EDIT2: Just one more question to the people using the custom kernel. Do you use the same initrd line as the stock kernel in menu.lst? Or can that be omitted, because everything is compiled in kernel?
Nope, no need for it
I'm currently working on the 2.6.27, I've already got it compiled and packaged, but this have to find time to test this and go through some minor version issues in the PKGBUILD.
Offline
If I'm not mistaken you can change a parameter of a running module by use of sysctrl. But if that doesn't work, post a message and I can rebuild this.
I have thought the model parameter affects the loading process of the module and can therefore not be changed at runtime. But I will try and report if it works.
thomasd wrote:EDIT2: Just one more question to the people using the custom kernel. Do you use the same initrd line as the stock kernel in menu.lst? Or can that be omitted, because everything is compiled in kernel?
Nope, no need for it
Just tried it, I get a kernel panic without the initrd line. The same panic as before with your kernel build that was not working.
Offline
Strange, I don't use any initrd whatsoever...
Offline
gothicknight wrote:If I'm not mistaken you can change a parameter of a running module by use of sysctrl. But if that doesn't work, post a message and I can rebuild this.
I have thought the model parameter affects the loading process of the module and can therefore not be changed at runtime. But I will try and report if it works.
"sysctl -a" does not list the parameter, and "sysctl -w kernel.pci.hda.snd-hda-intel.model=acer" says unknown key, as well as just "sysctl -w snd-hda-intel.model=acer".
Last edited by thomasd (2008-09-18 11:53:43)
Offline