Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
geek-sources (current:3.4.1)
View unanswered posts
View posts from last 24 hours

 
Reply to topic    Gentoo Forums Forum Index Unsupported Software
View previous topic :: View next topic  
Author Message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Sat Jun 09, 2012 6:28 pm    Post subject: geek-sources (current:3.4.1) Reply with quote

Just stumbled over the geek-sources from the init6-overlay

it spots and impressive set of patches e.g. UKSM(Ultra Kernel Samepage Merging) [http://www.youtube.com/watch?v=tQr0LQVn1os], the Barbershop Load Distribution algorithm (BLD) and other fixes and goodies


how many of you are using this kernel ?
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
188562
Apprentice
Apprentice


Joined: 22 Jun 2008
Posts: 186

PostPosted: Mon Jun 11, 2012 4:06 pm    Post subject: current 3.4.2 Reply with quote

yep i like it
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Mon Jun 11, 2012 5:59 pm    Post subject: Re: current 3.4.2 Reply with quote

init_6 wrote:
yep i like it


:D
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
188562
Apprentice
Apprentice


Joined: 22 Jun 2008
Posts: 186

PostPosted: Wed Jun 13, 2012 4:06 pm    Post subject: Reply with quote

At the moment sys-kernel/geek-sources include: AnotherUnionFS, Budget Fair Queueing Budget I/O Scheduler, BLD Alternate CPU load distribution technique for Linux kernel scheduler, Con Kolivas' high performance patchset, Spock's fbsplash, Fedora linux kernel patches, grsecurity, TuxOnIce, Intermediate Queueing Device patches, Mandriva/Mageia linux patches, Pardus linux patches, Ingo Molnar's realtime preempt patches, OpenSuSE linux patches and Ultra Kernel Samepage Merging patches

Also USE flag branding enable CONFIG_FONT_ISO_LATIN_1_8x16 - port of mac os x console font and the image of real penguin CONFIG_LOGO_LARRY_CLUT224

PS: All these features are controlled via USE flags
Back to top
View user's profile Send private message
Veldrin
Veteran
Veteran


Joined: 27 Jul 2004
Posts: 1945
Location: Zurich, Switzerland

PostPosted: Wed Jun 13, 2012 5:28 pm    Post subject: Reply with quote

I might even give it a try - as grsec patchset is included.
_________________
read the portage output!
If my answer is too concise, ask for an explanation.
Back to top
View user's profile Send private message
John R. Graham
Administrator
Administrator


Joined: 08 Mar 2005
Posts: 10587
Location: Somewhere over Atlanta, Georgia

PostPosted: Wed Jun 13, 2012 5:55 pm    Post subject: Reply with quote

Does "Con Kolivas' high performance patchset" include BFS? I assume so. Been wanting to check that out.

- John
_________________
I can confirm that I have received between 0 and 499 National Security Letters.
Back to top
View user's profile Send private message
188562
Apprentice
Apprentice


Joined: 22 Jun 2008
Posts: 186

PostPosted: Wed Jun 13, 2012 7:07 pm    Post subject: Reply with quote

The basic idea is to test each patch before applying. So any conflicting patches simply skipped. Yes, and geek-sources, there are no checks on the incorrect choice of mutually exclusive patches. "with great powers must come great responsibility"

And please report any bugs. I myself have been using and testing geek-sources… But they are so many opportunities to make mistakes.

Now i use sys-kernel/geek-sources with
Code:
equery u sys-kernel/geek-sources
[ Legend : U - final flag setting for installation]
[        : I - package is installed with flag     ]
[ Colors : set, unset                             ]
 * Found these USE flags for sys-kernel/geek-sources-3.4.2:
 U I
 + + aufs       : AnotherUnionFS.
 + + bfq        : Budget Fair Queueing Budget I/O Scheduler.
 + + bld        : Alternate CPU load distribution technique for Linux kernel scheduler.
 + + branding   : Enable Gentoo specific branding.
 - - build      : !!internal use only!! DO NOT SET THIS FLAG YOURSELF!, used for creating build images and the first half of bootstrapping
                  [make stage1]
 + + ck         : Enable Con Kolivas' high performance patchset.
 + + deblob     : Remove binary blobs from kernel sources to provide libre license compliance.
 - - fbcondecor : Use Spock's fbsplash patch.
 + + fedora     : Use Fedora kernel patches.
 - - grsecurity : Use grsecurity patches.
 + + ice        : Use TuxOnIce patches.
 - - imq        : Use Intermediate Queueing Device patches.
 + + mageia     : Use Mandriva/Mageia patches.
 + + pardus     : Use Pardus patches.
 - - reiser4    : Use Reiser4 FS patches.
 - - rt         : Use Ingo Molnar's realtime preempt patches.
 + + suse       : Use OpenSuSE patches.
 - - symlink    : Force kernel ebuilds to automatically update the /usr/src/linux symlink
 + + uksm       : Use Ultra Kernel Samepage Merging patches.
Back to top
View user's profile Send private message
khayyam
Watchman
Watchman


Joined: 07 Jun 2012
Posts: 6227
Location: Room 101

PostPosted: Sat Jun 16, 2012 5:46 pm    Post subject: Reply with quote

all ...

I spent the morning re-doing my .config using geek-sources (I didn't want to 'make oldconfig' and perhaps miss something, and so decided to make a clean start). I included BFS, BFQ, BLD, UKSM ... my primary goal being to see if anything could improve how the machine behaves under load. The machine in question is a Macbook 1,1 (circa 2006) and has an Intel GMA 950 integrated graphics controler. In short, if the machine is under any kind of load the display begins to re-draw, and seriously re-draw .. it takes about 10 seconds for the display to refresh, much like a curtain falling on a stage, completely un-usable ... a minute or so of this and the fan goes into vaccum cleaner mode. On OSX (which I'd ditched aeons back), the situation was even more dire, playing a low bitrate flv from youtube would cause such lag that you essencially loose control of the machine, the interface being soooo unresponsive it takes minutes to stop the process. I'd basically given up trying to use the machine while compiling, opting to blank the display and wait it out, or using nice, not an ideal situation for a gentoo install.

So, first impressions ... wow! ... I compiled the kernel twice in succession as a test, and while this was underway simply went about using the wm, tmux, vim, mutt, newsbuter, cmus, ff ... all without so much as a flicker on the display, yes the fan kicked in, but seemingly at at a lower RPM. In comparison with 3.2.12 and the CFQ scheduler, well, it just doesn't compare ... I coud use the machine under circumstances where I would have previously had to blank the display and go make a cup of tea.

I took ck's advice for non-tickless (CONFIG_NO_HZ=n) , CONFIG_HZ_1000=y, preempt, and disabling cgroups. Its difficult to factor in what part of this may be due to BFQ, BLD and UKSM but I imagine this is not entirely the result of the scheduler, so credit where credits due ...

Anyhow, you should probably give it a try ... just for the brain f**k :)

best ... khay
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Sat Jun 16, 2012 6:17 pm    Post subject: Reply with quote

@init_6:

I have a similar use-flag set to yours (no aufs & deblob)

3.4.1 wasn't usable for me due to usb keyboard not being initialized (probably all usb stuff)

3.4.2 works fine so far - ck-patchset but no bfs activated - there's ck2 available btw (ck1 still used in patchset)


alternatively instead of BFS you guys might want to give the RIFS Scheduler a try:

http://phoronix.com/forums/showthread.php?70954-The-v2-Rotary-Interactivity-Favor-Scheduler
http://code.google.com/p/rifs-scheduler/downloads/list


ck-patchset/bfs & rifs are mutually exclusive it seems

the creator of rifs also offers an improvement patch to BFS for kernel 3.3 - haven't tested if it works on top of bfs for 3.4



please consider adding following patch:

v2 libata: disable runtime pm for hotpluggable port_3.4-rc7-port

this makes sata hotplugging (eSATA) for me usable again - previously you could only attach one drive and when disconnecting/stopping the port is dead until system restart

Code:
diff -Nrupad linux-3.4-rc7/drivers/ata/ahci.c linux-3.4-rc7_no-reclaim_bak/drivers/ata/ahci.c
--- linux-3.4-rc7/drivers/ata/ahci.c   2012-06-02 01:11:18.244783741 +0200
+++ linux-3.4-rc7_no-reclaim_bak/drivers/ata/ahci.c   2012-05-22 01:38:38.944958848 +0200
@@ -1153,6 +1153,9 @@ static int ahci_init_one(struct pci_dev
    if (hpriv->cap & HOST_CAP_PMP)
       pi.flags |= ATA_FLAG_PMP;
 
+   if (hpriv->cap & HOST_CAP_SXS)
+      pi.flags |= ATA_FLAG_EXTERNAL;
+
    ahci_set_em_messages(hpriv, &pi);
 
    if (ahci_broken_system_poweroff(pdev)) {
diff -Nrupad linux-3.4-rc7/drivers/ata/ahci.h linux-3.4-rc7_no-reclaim_bak/drivers/ata/ahci.h
--- linux-3.4-rc7/drivers/ata/ahci.h   2012-06-02 01:11:18.245783741 +0200
+++ linux-3.4-rc7_no-reclaim_bak/drivers/ata/ahci.h   2012-05-22 01:38:38.999958848 +0200
@@ -172,6 +172,9 @@ enum {
    PORT_CMD_ALPE      = (1 << 26), /* Aggressive Link PM enable */
    PORT_CMD_ATAPI      = (1 << 24), /* Device is ATAPI */
    PORT_CMD_FBSCP      = (1 << 22), /* FBS Capable Port */
+   PORT_CMD_ESP      = (1 << 21), /* External SATA Port */
+   PORT_CMD_MPSP      = (1 << 19), /* Mechanical Presence Switch Attached to Port */
+   PORT_CMD_HPCP      = (1 << 18), /* Hot Plug Capable Port */
    PORT_CMD_PMP      = (1 << 17), /* PMP attached */
    PORT_CMD_LIST_ON   = (1 << 15), /* cmd list DMA engine running */
    PORT_CMD_FIS_ON      = (1 << 14), /* FIS DMA engine running */
diff -Nrupad linux-3.4-rc7/drivers/ata/libahci.c linux-3.4-rc7_no-reclaim_bak/drivers/ata/libahci.c
--- linux-3.4-rc7/drivers/ata/libahci.c   2012-06-02 01:11:18.245783741 +0200
+++ linux-3.4-rc7_no-reclaim_bak/drivers/ata/libahci.c   2012-05-22 01:38:39.033958852 +0200
@@ -1102,6 +1102,23 @@ static void ahci_port_init(struct device
    writel(1 << port_no, mmio + HOST_IRQ_STAT);
 }
 
+static bool ahci_port_pluggable(struct ata_port *ap)
+{
+   void __iomem *port_mmio = ahci_port_base(ap);
+   u32 cmd;
+
+   cmd = readl(port_mmio + PORT_CMD);
+
+   if ((ap->flags & ATA_FLAG_EXTERNAL) &&
+       (cmd & PORT_CMD_ESP))
+      return true;
+
+   if (cmd & (PORT_CMD_MPSP | PORT_CMD_HPCP))
+      return true;
+
+   return false;
+}
+
 void ahci_init_controller(struct ata_host *host)
 {
    struct ahci_host_priv *hpriv = host->private_data;
@@ -1117,6 +1134,9 @@ void ahci_init_controller(struct ata_hos
       if (ata_port_is_dummy(ap))
          continue;
 
+      if (ahci_port_pluggable(ap))
+         ap->flags |= ATA_FLAG_PLUGGABLE;
+
       ahci_port_init(host->dev, ap, i, mmio, port_mmio);
    }
 
diff -Nrupad linux-3.4-rc7/drivers/ata/libata-transport.c linux-3.4-rc7_no-reclaim_bak/drivers/ata/libata-transport.c
--- linux-3.4-rc7/drivers/ata/libata-transport.c   2012-06-02 01:11:18.245783741 +0200
+++ linux-3.4-rc7_no-reclaim_bak/drivers/ata/libata-transport.c   2012-05-22 01:38:39.031958855 +0200
@@ -292,9 +292,11 @@ int ata_tport_add(struct device *parent,
    }
 
    device_enable_async_suspend(dev);
-   pm_runtime_set_active(dev);
-   pm_runtime_enable(dev);
-   pm_runtime_forbid(dev);
+   if (!(ap->flags & ATA_FLAG_PLUGGABLE)) {
+      pm_runtime_set_active(dev);
+      pm_runtime_enable(dev);
+      pm_runtime_forbid(dev);
+   }
 
    transport_add_device(dev);
    transport_configure_device(dev);
diff -Nrupad linux-3.4-rc7/include/linux/libata.h linux-3.4-rc7_no-reclaim_bak/include/linux/libata.h
--- linux-3.4-rc7/include/linux/libata.h   2012-06-02 01:11:27.808783170 +0200
+++ linux-3.4-rc7_no-reclaim_bak/include/linux/libata.h   2012-05-22 01:37:48.374959331 +0200
@@ -187,6 +187,8 @@ enum {
    ATA_FLAG_SLAVE_POSS   = (1 << 0), /* host supports slave dev */
                    /* (doesn't imply presence) */
    ATA_FLAG_SATA      = (1 << 1),
+   ATA_FLAG_EXTERNAL   = (1 << 2), /* Controller supports external SATA */
+   ATA_FLAG_PLUGGABLE   = (1 << 3), /* Port is hotpluggable */
    ATA_FLAG_NO_ATAPI   = (1 << 6), /* No ATAPI support */
    ATA_FLAG_PIO_DMA   = (1 << 7), /* PIO cmds via DMA */
    ATA_FLAG_PIO_LBA48   = (1 << 8), /* Host DMA engine is LBA28 only */


thanks
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
khayyam
Watchman
Watchman


Joined: 07 Jun 2012
Posts: 6227
Location: Room 101

PostPosted: Sun Jun 17, 2012 8:58 am    Post subject: Reply with quote

kernelOfTruth wrote:
[...] 3.4.2 works fine so far - ck-patchset but no bfs activated - there's ck2 available btw (ck1 still used in patchset)

KoT, init6 ...

I moved geek-sources-3.4.2.ebuild into my local overlay, altered it to use patch-3.4.2-ck2, built (w/ DEFAULT_IOSCHED="bfq") and booted. So far no issues ...

best ... khay
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Sun Jun 17, 2012 3:52 pm    Post subject: Reply with quote

khayyam wrote:
kernelOfTruth wrote:
[...] 3.4.2 works fine so far - ck-patchset but no bfs activated - there's ck2 available btw (ck1 still used in patchset)

KoT, init6 ...

I moved geek-sources-3.4.2.ebuild into my local overlay, altered it to use patch-3.4.2-ck2, built (w/ DEFAULT_IOSCHED="bfq") and booted. So far no issues ...

best ... khay


thanks khay :)


I'll probably give that later a try


currently I'm trying out RIFS scheduler

there some minor issue so far:


when posting now parts of the remaining forum site are not updated when scrolling up and down (the lower part where all other posts are shown) via chromium

but that's only minor so far (and the only issue for far)


all I can say is that I'm pretty impressed by it - it feels more smooth and reactive than BFS

gotta have to give it a real stress-test now

:wink:


edit:

compiling libreoffice

rsyncing 1 TB of data to zlib-compressed btrfs

load at around 19-20

at the same time streaming 1080p HD youtube video - nice :D

all this with BLD (Barbershop Load Distribution algorithm) activated + RIFS


seems like I found my new favorite kernel ^^
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
Veldrin
Veteran
Veteran


Joined: 27 Jul 2004
Posts: 1945
Location: Zurich, Switzerland

PostPosted: Sun Jun 17, 2012 4:39 pm    Post subject: Reply with quote

I just had a brief look so far, but I am missing the configuration option for PaX and grsec, as there are in hardened sources.
Am I missing something, or is that a gentoo/hardened specific patch?

V.
_________________
read the portage output!
If my answer is too concise, ask for an explanation.
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Sun Jun 17, 2012 7:46 pm    Post subject: Reply with quote

Veldrin wrote:
I just had a brief look so far, but I am missing the configuration option for PaX and grsec, as there are in hardened sources.
Am I missing something, or is that a gentoo/hardened specific patch?

V.


most probably the patch didn't get applied so the option is missing

try a different use-flag combination


or a different reasno
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
khayyam
Watchman
Watchman


Joined: 07 Jun 2012
Posts: 6227
Location: Room 101

PostPosted: Mon Jun 18, 2012 5:06 am    Post subject: Reply with quote

kernelOfTruth wrote:
currently I'm trying out RIFS scheduler

I just noticed the patch was updated (about five hours ago) ... your right about BFS and RIFS being mutually exclusive, but having only done a --dry-run its difficult to say it they can co-exit or not, and with both patches being updated regularly it might be a headache to attempt to do so.

kernelOfTruth wrote:
all I can say is that I'm pretty impressed by it - it feels more smooth and reactive than BFS

Well, ck said that the idea that there can be a "one size fits all" scheduler is part of the problem. There is prediction involved and so its never any more accurate than a weather forcast. I've noticed that sometimes BFS does really well, then at other times it is as you say "less smooth" and more like a re-action than pro-action. If I compare it to CFQ then I can't say I'd want CFQ back, as I am having far less lag and a far more of usable interface while under any load. As I explained previously, with CFQ I might as well blank the display and say a prayer to the gods of 2982 processors. I should note that I'd removed PORTAGE_NICENESS while testing BFS and still have better interactivity while emerging, so BFS is providing some noticable effect.

Your "stress test" was pretty impressive, though I think your machine is somewhat more pokey than mine (full 1080p HD video is not possible with the GMA950, the display will go into 'curtaining').

kernelOfTruth wrote:
[...] all this with BLD (Barbershop Load Distribution algorithm) activated + RIFS

I too have BLD enabled, and I'll probably give RIFS a try at some point, btw did you patch geek-sources or was this some other kernel source?

best ... khay
Back to top
View user's profile Send private message
PaulBredbury
Watchman
Watchman


Joined: 14 Jul 2005
Posts: 7310

PostPosted: Mon Jun 18, 2012 7:13 am    Post subject: Reply with quote

khayyam wrote:
removed PORTAGE_NICENESS while testing BFS

That's a strange test.

A better test would be to do what users *should* be doing - give the scheduler correct hints about relative desired process responsiveness, rather than expecting the scheduler to read one's mind. As Con said.
Back to top
View user's profile Send private message
khayyam
Watchman
Watchman


Joined: 07 Jun 2012
Posts: 6227
Location: Room 101

PostPosted: Mon Jun 18, 2012 9:52 am    Post subject: Reply with quote

PaulBredbury wrote:
khayyam wrote:
removed PORTAGE_NICENESS while testing BFS

That's a strange test.

... "while", meaning, its not the test, I was reverting to a profile in which CFQ was unusable, it was simply for comparisons sake.

I'd read ck's comments re nice in the BFS FAQ and elsewhere, and yes, I understand that the scheduler isn't psychic, isn't this clear from what I said wrt "prediction"?

Thanks for the link re ionice though, currently the only application I'm ionice'ing is firefox ... and I wasn't aware of schedtool.

best ... khay
Back to top
View user's profile Send private message
PaulBredbury
Watchman
Watchman


Joined: 14 Jul 2005
Posts: 7310

PostPosted: Mon Jun 18, 2012 10:32 am    Post subject: Reply with quote

khayyam wrote:
I wasn't aware of schedtool

If you're testing BFS, and *not* using ionice -c3 schedtool -D for e.g. compilations, then it's a ridiculous test due to lack of proper configuring. The sort of half-assed test that Phoronix would perform ;)
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Mon Jun 18, 2012 10:44 am    Post subject: Reply with quote

khayyam wrote:
kernelOfTruth wrote:
currently I'm trying out RIFS scheduler

I just noticed the patch was updated (about five hours ago) ... your right about BFS and RIFS being mutually exclusive, but having only done a --dry-run its difficult to say it they can co-exit or not, and with both patches being updated regularly it might be a headache to attempt to do so.


if I understood correctly RIFS is a kind of modified, stripped and optimized CFS scheduler or something entirely new

so BFS and RIFS probably couldn't be activated both

there might be a way to make all three selectable but for the current state that's too much work (being that it changes every few days)

besides that I've lots of things to do

khayyam wrote:
kernelOfTruth wrote:
all I can say is that I'm pretty impressed by it - it feels more smooth and reactive than BFS

Well, ck said that the idea that there can be a "one size fits all" scheduler is part of the problem. There is prediction involved and so its never any more accurate than a weather forcast. I've noticed that sometimes BFS does really well, then at other times it is as you say "less smooth" and more like a re-action than pro-action. If I compare it to CFQ then I can't say I'd want CFQ back, as I am having far less lag and a far more of usable interface while under any load. As I explained previously, with CFQ I might as well blank the display and say a prayer to the gods of 2982 processors. I should note that I'd removed PORTAGE_NICENESS while testing BFS and still have better interactivity while emerging, so BFS is providing some noticable effect.

Your "stress test" was pretty impressive, though I think your machine is somewhat more pokey than mine (full 1080p HD video is not possible with the GMA950, the display will go into 'curtaining').


yeah, it's probably pretty beefy for desktop computer standards (core i7 860 quad core with HT and a radeon 5850 HD albeit with opensource graphics)

nevertheless it still got to a crawl (no smooth website scrolling, video watching, listening to sound, etc.) when under load and transferring data over and from the harddrive - this is sort of an infamous known issue with the amd64 architecture (see the amd64 subforum)

seems like there are several factors playing a role - and obviously the cpu scheduler, load balancer and i/o scheduler play an important role in this

the best test would be to newly transfer all the data from one harddrive to another and raise the pagecache, lower VFS pressure, etc. + at the same time do other multimedia-related stuff (several video streams, HD watching, audio from multiple sources) and heavy compiling

but currently I need to backup regularly & can't wipe the drives ^^

so I'll probably try that in a few weeks or so



khayyam wrote:
kernelOfTruth wrote:
[...] all this with BLD (Barbershop Load Distribution algorithm) activated + RIFS

I too have BLD enabled, and I'll probably give RIFS a try at some point, btw did you patch geek-sources or was this some other kernel source?

best ... khay


yeah I patched geek-sources and left ck-patchset un-selected - there are however a few patches which I'll try later, too (especially the swap-related ones)



the dev of the RIFS scheduler is active on the phoronix forums, btw:

http://phoronix.com/forums/showthread.php?70954-The-v2-Rotary-Interactivity-Favor-Scheduler/page11
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Mon Jun 18, 2012 10:59 am    Post subject: Reply with quote

there's an uksm related BUG-message which probably happened overnight while it was data-crawling (desktop file indexing) with Recoll (via xapian)

Quote:
[ 9528.160580] unrtf[30623] general protection ip:7f478e28fd76 sp:7fffa8ef1fd0 error:0 in unrtf[7f478e284000+16000]
[ 9529.295978] unrtf[30632] general protection ip:7fa2e986fd76 sp:7fff14346cb0 error:0 in unrtf[7fa2e9864000+16000]
[ 9535.580198] unrtf[30773] general protection ip:7f6caf751d76 sp:7fff3c315a90 error:0 in unrtf[7f6caf746000+16000]
[16242.457368] The following is only an harmless informational message.
[16242.457371] Unless you get a _continuous_flood_ of these messages it means
[16242.457373] everything is working fine. Allocations from irqs cannot be
[16242.457374] perfectly reliable and the kernel is designed to handle that.
[16242.457376] uksmd: page allocation failure. order:2, mode:0xc000
[16242.457378] Pid: 1253, comm: uksmd Not tainted 3.4.2-geek_btrfs+_libata_rifs_V3 #1
[16242.457380] Call Trace:
[16242.457393] [<ffffffff810043ae>] dump_trace+0x7e/0x300
[16242.457398] [<ffffffff8170d445>] dump_stack+0x69/0x6f
[16242.457403] [<ffffffff810e631e>] warn_alloc_failed+0xfe/0x1a0
[16242.457408] [<ffffffff810e919d>] __alloc_pages_nodemask+0x59d/0x810
[16242.457412] [<ffffffff810e94a2>] __get_free_pages+0x12/0x50
[16242.457417] [<ffffffff8111f460>] uksm_scan_thread+0x360/0x480
[16242.457422] [<ffffffff81073b45>] kthread+0x85/0x90
[16242.457428] [<ffffffff8171b4d4>] kernel_thread_helper+0x4/0x10
[16242.457431] Mem-Info:
[16242.457433] DMA per-cpu:
[16242.457434] CPU 0: hi: 0, btch: 1 usd: 0
[16242.457436] CPU 1: hi: 0, btch: 1 usd: 0
[16242.457437] CPU 2: hi: 0, btch: 1 usd: 0
[16242.457438] CPU 3: hi: 0, btch: 1 usd: 0
[16242.457440] CPU 4: hi: 0, btch: 1 usd: 0
[16242.457441] CPU 5: hi: 0, btch: 1 usd: 0
[16242.457442] CPU 6: hi: 0, btch: 1 usd: 0
[16242.457444] CPU 7: hi: 0, btch: 1 usd: 0
[16242.457446] DMA32 per-cpu:
[16242.457447] CPU 0: hi: 186, btch: 31 usd: 140
[16242.457449] CPU 1: hi: 186, btch: 31 usd: 180
[16242.457450] CPU 2: hi: 186, btch: 31 usd: 163
[16242.457451] CPU 3: hi: 186, btch: 31 usd: 186
[16242.457453] CPU 4: hi: 186, btch: 31 usd: 154
[16242.457454] CPU 5: hi: 186, btch: 31 usd: 70
[16242.457456] CPU 6: hi: 186, btch: 31 usd: 24
[16242.457457] CPU 7: hi: 186, btch: 31 usd: 31
[16242.457458] Normal per-cpu:
[16242.457459] CPU 0: hi: 186, btch: 31 usd: 61
[16242.457461] CPU 1: hi: 186, btch: 31 usd: 154
[16242.457463] CPU 2: hi: 186, btch: 31 usd: 37
[16242.457464] CPU 3: hi: 186, btch: 31 usd: 158
[16242.457466] CPU 4: hi: 186, btch: 31 usd: 166
[16242.457468] CPU 5: hi: 186, btch: 31 usd: 88
[16242.457469] CPU 6: hi: 186, btch: 31 usd: 183
[16242.457471] CPU 7: hi: 186, btch: 31 usd: 83
[16242.457474] active_anon:840795 inactive_anon:170269 isolated_anon:0
[16242.457475] active_file:440167 inactive_file:440201 isolated_file:0
[16242.457476] unevictable:0 dirty:276 writeback:0 unstable:0
[16242.457476] free:21009 slab_reclaimable:45140 slab_unreclaimable:37608
[16242.457477] mapped:19603 shmem:4324 pagetables:9285 bounce:0
[16242.457481] DMA free:15888kB min:60kB low:72kB high:88kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:15648kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:0kB slab_unreclaimable:16kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? yes
[16242.457486] lowmem_reserve[]: 0 2983 8023 8023
[16242.457492] DMA32 free:42688kB min:12160kB low:15200kB high:18240kB active_anon:1365552kB inactive_anon:344112kB active_file:574536kB inactive_file:574476kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:3055360kB mlocked:0kB dirty:52kB writeback:0kB mapped:3944kB shmem:0kB slab_reclaimable:76516kB slab_unreclaimable:34280kB kernel_stack:168kB pagetables:2988kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:34 all_unreclaimable? no
[16242.457498] lowmem_reserve[]: 0 0 5040 5040
[16242.457505] Normal free:25708kB min:20540kB low:25672kB high:30808kB active_anon:1997628kB inactive_anon:336964kB active_file:1186132kB inactive_file:1186328kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:5160960kB mlocked:0kB dirty:1052kB writeback:0kB mapped:74468kB shmem:17296kB slab_reclaimable:104044kB slab_unreclaimable:116136kB kernel_stack:3344kB pagetables:34152kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:6 all_unreclaimable? no
[16242.457511] lowmem_reserve[]: 0 0 0 0
[16242.457513] DMA: 0*4kB 0*8kB 1*16kB 0*32kB 2*64kB 1*128kB 1*256kB 0*512kB 1*1024kB 1*2048kB 3*4096kB = 15888kB
[16242.457521] DMA32: 7691*4kB 1413*8kB 0*16kB 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 42068kB
[16242.457528] Normal: 5506*4kB 163*8kB 11*16kB 3*32kB 1*64kB 1*128kB 1*256kB 1*512kB 1*1024kB 0*2048kB 0*4096kB = 25584kB
[16242.457535] 885751 total pagecache pages
[16242.457537] 1024 pages in swap cache
[16242.457538] Swap cache stats: add 7731, delete 6707, find 1/2
[16242.457539] Free swap = 9410356kB
[16242.457540] Total swap = 9437180kB
[16242.485901] 2097136 pages RAM
[16242.485903] 58102 pages reserved
[16242.485904] 544219 pages shared
[16242.485905] 1591927 pages non-shared
[16271.219009] BUG: unable to handle kernel NULL pointer dereference at 0000000000000070
[16271.219044] IP: [<ffffffff8111cbde>] uksm_do_scan+0x5ee/0x2b10
[16271.219069] PGD 0
[16271.219080] Oops: 0000 [#1] PREEMPT SMP
[16271.219105] CPU 4
[16271.219112] Modules linked in: radeon drm_kms_helper ttm drm cfbfillrect cfbimgblt cfbcopyarea i2c_algo_bit it87 hwmon_vid nf_conntrack_ftp nf_conntrack_irc ipt_REJECT xt_state xt_LOG xt_limit iptable_filter snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_page_alloc snd_timer snd e1000e soundcore i2c_i801 wmi coretemp i7core_edac e1000 lockd auth_rpcgss sunrpc scsi_wait_scan sl811_hcd ohci_hcd ssb usb_storage ehci_hcd [last unloaded: firewire_core]
[16271.219379]
[16271.219386] Pid: 1253, comm: uksmd Not tainted 3.4.2-geek_btrfs+_libata_rifs_V3 #1 Packard Bell ipower G3710/FMP55
[16271.219428] RIP: 0010:[<ffffffff8111cbde>] [<ffffffff8111cbde>] uksm_do_scan+0x5ee/0x2b10
[16271.219457] RSP: 0018:ffff88023196dd90 EFLAGS: 00010286
[16271.219475] RAX: 0000000000000000 RBX: ffff88023526b800 RCX: 0000000000000000
[16271.219498] RDX: ffff88023526b800 RSI: ffff88023605aa70 RDI: 0000000000000000
[16271.219521] RBP: ffff8801055187e8 R08: ffff88023196c000 R09: ffffffff81fa97d0
[16271.219544] R10: 0000000000000000 R11: 0000000000000000 R12: ffff880236b12e00
[16271.219567] R13: 0000000000000000 R14: ffff880232967a50 R15: ffff880132ffe000
[16271.219590] FS: 0000000000000000(0000) GS:ffff88023fd00000(0000) knlGS:0000000000000000
[16271.219617] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[16271.219636] CR2: 0000000000000070 CR3: 000000020e70a000 CR4: 00000000000007e0
[16271.219667] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[16271.219704] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[16271.219743] Process uksmd (pid: 1253, threadinfo ffff88023196c000, task ffff88023605aa70)
[16271.219786] Stack:
[16271.219799] ffff88023605aa70 0000000000000000 ffff8800054a6b00 ffffea0008c143c0
[16271.219847] ffffea0004fac900 ffff880215c63798 ffffffff81e281c0 0000000000000001
[16271.219891] ffff880236b12e00 0000000081719492 ffff88023196dfd8 ffffffff817180ba
[16271.219924] Call Trace:
[16271.219939] [<ffffffff8111f4c9>] uksm_scan_thread+0x3c9/0x480
[16271.219962] [<ffffffff81073b45>] kthread+0x85/0x90
[16271.219983] [<ffffffff8171b4d4>] kernel_thread_helper+0x4/0x10
[16271.220003] Code: 24 18 02 48 8b 45 00 49 39 c4 0f 84 45 14 00 00 48 8b 50 70 48 8b 12 48 39 da 75 18 48 8b 08 49 39 cc 0f 84 2a 14 00 00 48 89 c8 <48> 8b 49 70 48 39 11 74 e8 48 89 c5 e9 22 14 00 00 48 8b 7c 24
[16271.220267] RIP [<ffffffff8111cbde>] uksm_do_scan+0x5ee/0x2b10
[16271.220289] RSP <ffff88023196dd90>
[16271.220302] CR2: 0000000000000070
[16271.242656] ---[ end trace c44f6862bcdebc40 ]---

_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
PaulBredbury
Watchman
Watchman


Joined: 14 Jul 2005
Posts: 7310

PostPosted: Mon Jun 18, 2012 11:05 am    Post subject: Reply with quote

kernelOfTruth wrote:
when under load and transferring data over and from the harddrive

Other things to try:

setpci:
Code:
    # Raise priority of audio card
    # https://wiki.archlinux.org/index.php/Pro_Audio
    setpci -d *:* latency_timer=b0
    setpci -s 00:1b.0 latency_timer=ff
    # Video card gets next highest
    setpci -s 01:00.0 latency_timer=d0

And check interrupts.
Back to top
View user's profile Send private message
khayyam
Watchman
Watchman


Joined: 07 Jun 2012
Posts: 6227
Location: Room 101

PostPosted: Mon Jun 18, 2012 11:28 am    Post subject: Reply with quote

PaulBredbury wrote:
If you're testing BFS, and *not* using ionice -c3 schedtool -D for e.g. compilations, then it's a ridiculous test due to lack of proper configuring. The sort of half-assed test that Phoronix would perform

You seem to be labouring under the impression that "test" means more than "gathering an inital impression", and that I'm running a sheduler test facility.

Indeed it's "ridiculous" to make a change in something and then to contrast that change with a prior state of affairs ... but such ridiculousness is exactly the level at which I was operating, too bad for me I'm so "half-assed".

Are you usually so trollish?

best ... khay
Back to top
View user's profile Send private message
kernelOfTruth
Watchman
Watchman


Joined: 20 Dec 2005
Posts: 6111
Location: Vienna, Austria; Germany; hello world :)

PostPosted: Mon Jun 18, 2012 11:43 am    Post subject: Reply with quote

PaulBredbury wrote:
kernelOfTruth wrote:
when under load and transferring data over and from the harddrive

Other things to try:

setpci:
Code:
    # Raise priority of audio card
    # https://wiki.archlinux.org/index.php/Pro_Audio
    setpci -d *:* latency_timer=b0
    setpci -s 00:1b.0 latency_timer=ff
    # Video card gets next highest
    setpci -s 01:00.0 latency_timer=d0

And check interrupts.


thanks a lot Paul !

been a long time (a few years) I last time tried this


if I remember correctly USB is especially badly affected (usb keyboard + mouse) - PS/2 wasn't making that much trouble

don't know if it works but raising the priority of USB also might help

good stuff ! :)
_________________
https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa

Hardcore Gentoo Linux user since 2004 :D
Back to top
View user's profile Send private message
188562
Apprentice
Apprentice


Joined: 22 Jun 2008
Posts: 186

PostPosted: Mon Jun 18, 2012 5:08 pm    Post subject: Reply with quote

kernelOfTruth wrote:
alternatively instead of BFS you guys might want to give the RIFS Scheduler a try:

http://phoronix.com/forums/showthread.php?70954-The-v2-Rotary-Interactivity-Favor-Scheduler
http://code.google.com/p/rifs-scheduler/downloads/list


ck-patchset/bfs & rifs are mutually exclusive it seems

the creator of rifs also offers an improvement patch to BFS for kernel 3.3 - haven't tested if it works on top of bfs for 3.4


thx. maybe in the next version will be added.

kernelOfTruth wrote:
please consider adding following patch:

v2 libata: disable runtime pm for hotpluggable port_3.4-rc7-port

this makes sata hotplugging (eSATA) for me usable again - previously you could only attach one drive and when disconnecting/stopping the port is dead until system restart


patch v2 libata: disable runtime pm for hotpluggable is not applied even on vanilla kernel :(



Now I'll add sys-kernel/geek-sources-3.4.3 with a new set of patches from PLD Linux http://cvs.pld-linux.org/cgi-bin/viewvc.cgi/cvs/packages/kernel/?pathrev=MAIN

pld USE flag is masked because of all the patches normally are applied only a few.

I think to do to fix them and adapting. For example there is a kernel-reiser4.patch and it is for kernel 3.4.3!!!
Back to top
View user's profile Send private message
tclover
Guru
Guru


Joined: 10 Apr 2011
Posts: 516

PostPosted: Wed Jun 20, 2012 12:55 pm    Post subject: thanks for sharing! Reply with quote

Well, because of this thread I noticed that BFQ was/is not dead because I lost track of its development/release after the months needed to release v3x after kernel 3.x was released.

Although it was not simple to add that patch to my git-sources (not geek ;-) ebuild because of the missing *.patch.bz2 archive, I finally added it (checkout my sig.).

Now, I'm very curious about BLD or RIFS[-ES] patch set and I'd gladly add them to my git-sources:3.4.x ebuild if it wasn't in such alpha testing phase... trying to avoid hours of ebuild maintenance.
_________________
home/:mkinitramfs-ll/:supervision/:e-gtk-theme/:overlay/
Back to top
View user's profile Send private message
dusanc
Apprentice
Apprentice


Joined: 19 Sep 2005
Posts: 248
Location: Serbia

PostPosted: Wed Jul 04, 2012 8:18 am    Post subject: Reply with quote

Quote:
....For example there is a kernel-reiser4.patch and it is for kernel 3.4.3!!!

I saw that too, have to dig a little and see what modifications are going on there....

PS. based on R4 for 2.6.35?
_________________
Reiser4 Gentoo FAQ [25Dec2016]
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Unsupported Software All times are GMT
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum