/dev/chat

Ramesh

00:01

@slm, I still remember how you helped me out when I was trying to setup RAID in one of my lab machines.

2

I am a fairly new systems administrator and I have to configure the RAID on a Linux machine. From my understanding so far, RAID is an array of multiple disks. I have some questions though. Question 1: If I have five 500 GB HDD, these 5 HDDs together are termed as RAID disk.Is this right? Quest...

:)

slm

00:15

@Ramesh Once you get past the terminology setting up a RAID is actually one of the easier things to do. Now that you've done it. Wouldn't you agree? 8-)

Ramesh

@slm, yes you are correct :)

Braiam

00:35

who migrated this? unix.stackexchange.com/posts/118237/revisions

slm

Says Michael did it

Braiam

...

Ubuntu Ultimate isn't Ubuntu despite the name :/

slm

> Post Migrated Away to askubuntu.com by Michael Mrozek♦

derobert

04:06

@Webby you around? I see you have a new question...

doing primary & secondary in parallel is easy: put an & at the end of the command to run it in the background, then put a wait at the end of the script to make sure its done before the script exits.

Also, your script doesn't do the same thing as before—it copies all the files from B, then it copies all the files from C, if they're either different or missing.

Webby

04:31

Yeah I am around

Yeah it will copy only the files from B which I am passing from an array

and some files won't be there in B

so then from that same array, it will copy the files from C

as I mention, files will be unique in B and C

Still something is wrong?

@mikeserv: are you around?

mikeserv

Yeah. Now I am. Gotta turn this pager thing off...

Webby

hey

tell me one thing

mikeserv

Hi.

Sure.

Webby

@derobert mention that my logic is wrong?

mikeserv

I dunno.

Did he?

Webby

04:41

I am trying to copy files from machineB, if it is not there then I need to copy that file from machineC

see his comment above

mikeserv

Yeah.

Ok.

Webby

He said - Also, your script doesn't do the same thing as before—it copies all the files from B, then it copies all the files from C, if they're either different or missing.

mikeserv

If they're different or missing depends on the options. By default rsync will first use it's own engine - hashes, whatever. You use -?

a?

I forget?

Webby

machineB and machineC will have unique files

if total files are 100

then machineB can have 40 and machineC may have 60

mikeserv

OK. I mean the operands, rsync ???

I get it - primary, secondary, yes?

Webby

04:45

yeah

so in my case if PRIMARY files are not there in machineB

mikeserv

If preferred and otherwise. Ok. so what are rsyncs operands?

B > C. Understood.

Webby

yeah

which operands?

mikeserv

% rsync ??? src:file tgt:file

???

Webby

I have no idea.. need to check

mikeserv

Ok.

I know that before it was doing a file at a time. If @derobert has altered it and avoided use of wholefile or delete-target or something, then it will use some differencing mechanism.

So long as it is a continuous process. Without a daemon and/or specific arguments rsync won't necessarily avoid updating an up-too-date file. But if it's given two sources and one target it should do with little config.

Definitely -t is a good way to ensure it's doing the right thing - especially if you're working with large database files as you seem to be.

Must be a lot of operands.

Webby

04:52

where I would be adding -t

rsync -tavz

like this?

mikeserv

With the others. Let me make sure. V should be verbose, a for all. That might not be right for two different sources - if one file is owned by some other user or something it might not register as the same as one that's already there.

Are your file permissions pretty closely handled?

Webby

yeah

mikeserv

I mean - you've got StrictHostKeyChecking no set when logging into a datacenter so I've gotta ask that one..

Webby

I added to avoid the issue when you ssh it give RSA signature thing

right

mikeserv

So b/file1 is guaranteed to differ only by access/modification time from c/file1?

Well, I won't tell anybody if you don't.

Webby

04:56

in b files will be different altogether as compared to in c

mikeserv

If that's the case then all of the above should be fine and -t is not needed. Yeah, different as in newer, right?

Webby

by different I mean

in B and C total number of files will be 1680

mikeserv

Not different owners or file attributes. But changed.

Right?

Webby

yeah

so B will have around 700 files and C will have around 800 files

mikeserv

Ok. Then you should be set. -t isn't necessary, but it will avoid unnecessary file hashing when it can.

Webby

04:59

And B will have unique files and C will also have unique files

mikeserv

Well, man the best I can say is to do this: use -n AND... just a script option...

Webby

there won't be any duplicate files

mikeserv

Ok. Test it, man. --list-only

See what it does.

Where is the script now? Is it on this site?

Webby

yeah on the same question

you commented out

mikeserv

That was current? There's like 4 of em - I'm sorry.

Webby

05:02

0

Q: How to parallelize the rsync while copying the files?

I am running my below shell script from machineA which is copying the files from machineB and machineC into machineA. If the files are not there in machineB, then it should be there in machineC. Files won't be duplicated across machineB and machineC, they will have unique files so let's say I hav...

linux bash shell rsync parallel

mikeserv

Why is it still deleting stuff?

Webby

before copying files in machinA in primary directory, I need to delete the old files

and then copy new files

same with secondary

mikeserv

Well it won't be updating deleted files, that's for sure.

Webby

meaning?

New files are there in machineB and C

I just need to copy those new files in machineA

so I will delete old files from machineA first

and then copy new files

mikeserv

So that's rsync's job. It compares old files with new files and copies as little as possible to change the old file to match the new file. That's it's primary purpose.

But I guess you're doing that fresh once per week?

Webby

05:06

hmmm

twice per week

mikeserv

Ok, well, as is, I don't believe C wil overwrite B as long as B is newer than C.

If a C file is newer than B I think it might - but that's a minor thing that you need to verify with tests.

If you can work out a strategy to only rsync without deleting, depending on the actual file turnover, you could notice a very significant transfer speed decrease.

But if it were me, I would be doing it in a tiny KVM subnet.

And fake files.

Webby

I see

problem is we dont have fake files :)

only real files

as of now

mikeserv

05:24

Of course, but, you know : for f in $(seq 1 100) ; do touch ./f ; done

Now you have fake files.

And touch updates modification times. So you can know what will happen when one is newer than another.

Failing that though... rsync avz --list-only

The above should read "transfer time decrease" ... kinda funny.

slm

06:04

@mikeserv - you can make the files like this too: touch {1..100}

mikeserv

Oh, well probably, yeah.

Hey, man, is this common? Like, do you think that guy even knows what that does? And he's gonna go trust two corporate datacenters to it?

Graeme

09:58

Is this guy asking what I think he is asking?

0

Q: How to configure proxy settings over the terminal?

I am able to configure an open proxy over the Firefox in windows and backtrack but i am not able to configure it in the terminal in backtrack.

terminal proxy backtrack

Faheem Mitha

11:52

I just noticed that dmesg has this message repeated over and over many times. Any idea why, and how can it be removed?

[4291386.788100] usb 3-2: reset high-speed USB device number 2 using ehci_hcd
[4291386.921750] usblp0: USB Bidirectional printer dev 2 if 0 alt 0 proto 2 vid 0x04E8 pid 0x328D
[4291507.224516] usblp0: removed
[4291517.880553] usblp0: USB Bidirectional printer dev 2 if 0 alt 0 proto 2 vid 0x04E8 pid 0x328D
[4291517.880760] usblp0: removed

I've got a usb printer and it works fine.

slm

@mikeserv Don't know.

terdon

17:49

@slm unix.stackexchange.com/questions/120898/…

OP took your advice, might want to delete the comment.

slm

18:14

@terdon - thanks deleted.

slm

19:21

Isn't this off topic? It's about PuTTY SSH problems?

1

Q: SSH agent forwarding, multiple hop

I'm trying to set up SSH agent forwarding with multiple hops. My local machine is Windows, with putty and pageant. The problem is that my private keys, originating from windows, work only on the first hop. Private key added on any other machine works everywhere. First hop machine is called 'vpn...

ssh putty ssh-agent

terdon

19:51

> My local machine is Windows, with putty and pageant.

I'd say that makes it off topic yeah.

Gilles

@slm uh? Unless the server is also running Windows, it's on-topic here

interoperability between unix and non-unix stuff (Windows, hardware, etc.) is on-topic

slm

@Gilles - that's why I asked, still seems off topic to me

Gilles

we've accepted questions about putty for years

as long as the problem is related to using it to reach a unix machine, e.g. ssh key issues

not if the problem was between putty and windows of course

slm

@Gilles - I already voted to close it when terdon agreed.

it now has 3

terdon

@Gilles is it? Even if the solution is fixing a configuration on the Windows side?

Gilles

20:11

@terdon in this case, yes. We've got plenty of similar questions

just about anything tagged putty, for example

terdon

@Gilles OK, I'll write a meta post when I get the chance, this should be clarified.

@slm retracted my close vote, can you do the same? Maybe not 'cause yours was the first

Gilles

20:27

@terdon I'm surprised that it hasn't come up before on meta, but I can't find anything relevant

terdon

Neither could I, at least not with a quick search.

Faheem Mitha

@mikeserv if you are Prince William I want to hear about the wedding.

mikeserv

Marriage is for suckers. Just ask my exwife.

Faheem Mitha

@mikeserv :-)

What is this nvidia-prime package on Ubuntu?

Gilles

I did find this question for the very early beta whose most-upvoted answer goes contrary to the stance the site took:

7

Q: Is it On or Off Topic When Unix/Linux is Only Ancillary to the Question?

This is not necessarily a "specific question" discussion. It's just an example to raise the issue. I'm looking at this question: Good introduction to SubVersion for unexperienced users? This is a tough one; deciding the scope of your site. This question is about SubVersion... which is running on...

discussion on-topic

terdon

20:57

@Gilles why the mention of cross-platform? The issue here is not if a version exists that runs on different platforms but whether you are running it on a different platform. I think your discussion of cross platform is misleading.

Braiam

anyone got an idea why "mv: cannot move ‘/sbin/ifquery’ to ‘/sbin/ifquery.bk’: Input/output error" appear?

Gilles

@terdon this has nothing to do with our discussion here, it's just something unrelated I came upon while searching for a relevant meta post

Braiam

filesystem is ok

Gilles

@Braiam with on-disk filesystems, I/O error is usually a driver bug or hardware problem

what filesystem?

anything in the kernel logs?

Braiam

mm... wait

Gilles

21:07

could also be a miscellaneous error with a non-disk filesystem

Braiam

it's a locked VM, so I'm not surprised they fixed some files, the dmesg is empty

slm

21:22

@terdon retracted

@terdon I couldn't either, that's why these Q's seemed off topic to me as well.

Faheem Mitha

21:33

My X server was using 1.9 Gig RSS and like 50% CPU when I wasn't doing anything. Currently at 335 Mb. I logged out, restarted X, and then logged back in again. I was having problems with video playback. Even the keyboard seemed unresponsive. My quess is that Chromium is the reason. Is this a memory leaks issue, or something else? Any ideas?

I don't know if a question like this is suitable for the site - it is pretty vague.

Braiam

@FaheemMitha there is any log?

Faheem Mitha

@Braiam log where?

strugee

@FaheemMitha if you're having driver problems I feel bad for you son. I got 99 problems but a blob ain't one

aka: are you using proprietary drivers?

Faheem Mitha

@strugee indeed i am. nvidia.

strugee

try nouveau and see if the leaks go away

Faheem Mitha

21:36

@strugee hmm. wonder if my card is supported. do you recommend nouveau? does it support two screens? and is it a leak or something else?

strugee

@FaheemMitha no idea. I don't think I've ever used GNU/Linux on an Nvidia card

you can always get rid of it later

Faheem Mitha

@strugee Ok. What card do you use?

strugee

@FaheemMitha various

IIRC the only Nvidia card that I have is on a MacBook with a dual-chip deal. integrated Intel, dedicated Nvidia. never got the Nvidia card to work.

mikeserv

@Braiam the filesystem may appear rw to VM but be ro on host. That would make for an io error -especially if its exhausted its cowspace already.

strugee

I'll do some tinkering when I finally switch it over to Arch (probably this afternoon)

Faheem Mitha

21:41

@strugee maybe i'll try it some time. right now I don't feel like breaking my system

strugee

@FaheemMitha nods

Braiam

mount says aufs on / type aufs (rw)

mikeserv

@Braiam Yeah - so what about cp? MV will present some problems with layered filesystems.

Faheem Mitha

i think someone (@derobert?) suggested using an Intel video card. Don't remember any details. I don't want an onboard card, though. Hate those things.

Braiam

well, I need mv, since I want to replace it... so that's not going to work

strugee

21:49

@FaheemMitha that would be because Intel is badass and we love them for their GNU/Linux support

like, there is no proprietary Intel GPU driver because they work directly on the open-source one

Faheem Mitha

@strugee right, but someone give me a Intel video card recommendation, then

Braiam

@FaheemMitha or you would like ati...

Faheem Mitha

@Braiam eww

Braiam

I take an OOTB working ati than a nvidia awesomeness after messing with it for a week

Faheem Mitha

@Braiam ootb?

Braiam

21:51

out of the box

Faheem Mitha

@Braiam ah

so. anyone using an Intel card? Talk to me.

mikeserv

@Braiam, if you want to replace it, bind mount the the replacement file over it.

That's as much as aufs could do anyway.

@FaheemMitha an Intel card? Do they make those?

Faheem Mitha

@mikeserv So I heard.

i'm using an amd platform. i wonder if that is a problem

mikeserv

I've never heard of em. Thats pretty cool. My laptop uses only the Sandy Bridge on-chip acceleration as it lacks discrete graphics, and its not awful. Though its mostly only at a console anyway.

strugee

@mikeserv I don't think Intel does dedicated, only integrated

mikeserv

22:05

My desktop is AMD - i like it, but I'm cheap.

Oh.

strugee

yeah

Faheem Mitha

it is possible intel doesn't do standalone graphics cards. i can't find any mention of them

which is a deal killer with my current system

derobert

@FaheemMitha Unfortunately, as far as I know, you can only get Intel graphics on-board

Faheem Mitha

@derobert yes, that is what i thought. are there are other alternatives to nvidia and ati?

mikeserv

My first desktop had a Diamond Viper 8mb VESA card - it could do 16.7million colors. Kings Quest never looked better.

strugee

22:08

not really

mikeserv

Really.

derobert

@FaheemMitha I think there may be a few... but not really.

Faheem Mitha

@derobert recall i was complaining about mplayer? looks like the problem was with X. i restarted the server, and things are back to normal, more or less.

@derobert bummer

strugee

@mikeserv sorry. I was replying to @FaheemMitha. race conditions may apply.

derobert

Non-current-generation nVidia cards are supported by nouveau fairly well, I believe

Faheem Mitha

22:09

@strugee do the reply thingy.

mikeserv

I know. It was just fun to say.

derobert

Same is probably true with ATI.

strugee

@FaheemMitha I know, I know...

:P

Faheem Mitha

@derobert yes, that is what @strugee suggested. i've not been brave enough to try

mikeserv

Yeah -up to Kepler nouveau is very good. Especially since 3.11

Faheem Mitha

22:10

@derobert can i get something like twin view?

strugee

@FaheemMitha it's not that nouveau is harder to use, it just isn't quite as fast

Faheem Mitha

I have GeForce 7300 SE/7200 GS. not exactly a new card.

derobert

@FaheemMitha dual-display? You should be able to. I haven't used an ATI card in a while, so I can't speak from experience...

Faheem Mitha

i can't find any mention of how nouveau works with it

mikeserv

No. Nouveau will handle that very well.

strugee

22:10

it's more stable than the Nvidia proprietary blob, which is why I suggested it. much less likely to have bugs

Faheem Mitha

@mikeserv are you replying to me?

@derobert nvidia, not ati

mikeserv

There's a data table on their website that's discusses card family capabilities.

@FaheemMitha yes

Faheem Mitha

@mikeserv i saw something. i didn't understand it. how do i know what family my card belongs to?

@mikeserv reply thingy.

derobert

@FaheemMitha I was only driving a single display of the multi-port card at work, but it should work. And its a newer card than that 7300...

Faheem Mitha

@derobert which card?

derobert

22:12

03:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 440] (rev a1)

mikeserv

@FaheemMitha its different on the phone.

Faheem Mitha

@mikeserv ok

mikeserv

Sorry.

Faheem Mitha

@derobert ok. maybe i'll give it a try.

@mikeserv so you think the nouveau should work with my card, including dual display?

mikeserv

Here: nouveau.freedesktop.org/wiki/FeatureMatrix

derobert

22:14

And, except for the signal mask issue, the nVidia proprietary drivers have been quite stable for a while...

mikeserv

Yes.

Faheem Mitha

@mikeserv yes, that is what i saw. but i've no idea where my card is in there

@derobert signal mask issue?

derobert

@FaheemMitha Recent bug they had... let me go find the question again

3

Q: After upgrade, X button in titlebar no longer closes xterm

In theory, pressing the close button should close an xterm. It did before I did an aptitude upgrade. (I'm running Debian testing). But now it doesn't—clicking the X does nothing—but only for some xterms. In particular, if I launch an xterm from the KDE quicklaunch thing, or the K menu, those xt...

linux debian x11 xterm kwin

... that bug.

mikeserv

@derobert True, but that thing they did recently where they had to hack in 32-bit code to their driver to keep their ROMs private - again, since 3.11 - really ticked me off. Then again, little Michael likes cartoons and I can't stomach both a cable and internet bill, so I'm locked in I guess.

Faheem Mitha

looks like my card is the NV40 family. can anyone confirm?

derobert

22:19

@mikeserv what was that they did? I didn't hear about it...

Faheem Mitha

@derobert ok, i vaguely remember you filing that

is this for dual head support?

nouveau.freedesktop.org/wiki/Randr12

derobert

yes, that's dual-head support

mikeserv

@FaheemMitha yes nv40 nouveau.freedesktop.org/wiki/CodeNames

derobert

or configure at runtime with xrandr. If your DE doesn't do it for you

mikeserv

@derobert I'll try to find it

Faheem Mitha

22:21

Doesn't look too bad. Maybe I'll give it a try. it suggests that maybe i don't actually need to do any configuration. which is how i like it.

derobert

yeah, it should just work

Braiam

mm... interesting askubuntu.com/q/437764/169736

derobert

the only configuration you'd need to do is if it puts your monitors in the wrong order, or you want to use non-default resolutions.

mikeserv

devtalk.nvidia.com/default/topic/628864/…

Faheem Mitha

@derobert ok. that does sound encouraging.

mikeserv

22:22

If they do get in the wrong order you can specify unique names with kernel parameters

Faheem Mitha

@mikeserv hmm. at boot time?

derobert

here is my entire monitor config on an Intel card... Nouveau is about the same.

Faheem Mitha

@derobert this thing about X mem usage going to 2 gigs. might that be memory leakage or something else?

mikeserv

Yeah. Even ... Darn is it eid? I forget what that thing is called... Tip of typing finger...

derobert

Section "Monitor"
        Identifier "Left"
EndSection

Section "Monitor"
        Identifier "Right"
        Option "RightOf" "Left"
EndSection

Section "Device"
        Identifier "Sandybridge"
        Driver "intel"
        Option "monitor-HDMI2" "Right"
        Option "monitor-HDMI1" "Left"
        Option "AccelMethod" "sna"
EndSection

Faheem Mitha

22:23

@derobert short and sweet. this does dual-head automatically?

strugee

your DE should also be able to help you configure that stuff

Faheem Mitha

@strugee ok. i mostly ignore KDE though

I'm a bit encouraged. maybe i'll give it a try in the next day or two

mikeserv

No, that was double-talk for - we don't want the Linux kernel to have access to our BIOS memmapped space so we're wrapping our driver in 32-bit code to protect high mem

derobert

@FaheemMitha yes, that's dual head

Faheem Mitha

@derobert thoughts about the memory usage?

derobert

22:26

@mikeserv what's interesting about that???

strugee

@derobert presumably that Nvidia is an asshole of a company

Faheem Mitha

@Braiam i don't think debian supports anything like that

derobert

Am I missing something in that linked post?

@FaheemMitha I haven't looked at xorg memory usage in a long time. I can only suggest xrestop

Braiam

@FaheemMitha I know, it may need a hack in whatever script he uses

derobert

@mikeserv how did you read that out of the num_physpages post????

mikeserv

22:28

What's interesting? Well, nvidia is retroactively hamstringing your video cards capabilities and adding unnecessary dependencies to protect their right to run their own code unmolested in your memory.

Faheem Mitha

@Braiam i think it would require debian support. otherwise how can i know what updates to poll for?

@derobert i have no idea what xrestop is, but i'll try it. thanks

derobert

@FaheemMitha xrestop tells you how much X server memory your X clients are using

Faheem Mitha

@derobert that sounds handy. thanks

mikeserv

Also, when they did it, it was kind of a rushed thing. It broke nvidia for many for a few months at least. And they didn't even respond to the months' old for posts on their own forums.

Faheem Mitha

apparently something called kwin is using a ton of memory. what is that?

derobert

22:29

@FaheemMitha window manager

mikeserv

The same one I linked you to.

Faheem Mitha

@derobert oh

mikeserv

I apologize if its boring, but it seemed on topic.

derobert

@mikeserv the one you linked to talks about working around a kernel ABI change...

because it sounds like some of their cards don't support 64-bit DMA

mikeserv

Yes. The reason they do not support it is what is interesting.

derobert

22:32

The reason the cards don't support 64-bit DMA?

I'd assume because nothing did?

mikeserv

Seriously? Do you mean nothing nvidia?

derobert

I mean, the nvidia drivers support a lot of older hardware, some of it PCI. Practically no PCI cards or controllers supported 64-bit addressing. Early PCI-e didn't, either....

Even on my fairly new computer here, half of the on-board PCI/PCI-e devices are 32-bit.

mikeserv

Yes. This is true. And this is a very good reason to offer two drivers - which they do.

derobert

They don't offer two. They have an older, no-longer-updated version for truely ancient cards

mikeserv

Yes, they do. Nvidia legacy and nvidia. And this is not to mention their ever present beta and alpha offerings.

And it is updated - nvidia legacy. It is updated as older cards are subsumed by it.

google.com/…

derobert

22:41

So, the current drivers go back to cards that are newer than 2006 or so

Did AGP even support 64-bit addressing?

mikeserv

No, never that.

derobert

Presumably, they just changed the way they handle the older cards with 32-bit address support. Does that code get used at all on 64-bit capable cards?

03:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 440] (rev a1) (prog-if 00 [VGA controller])
        Subsystem: ASUSTeK Computer Inc. Device 83b7
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
        Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
        Latency: 0
        Interrupt: pin A routed to IRQ 94
        Region 0: Memory at fa000000 (32-bit, non-prefetchable) [size=16M]

I'm guessing that '64-bit' in there means the newer one does handle 64-bit addressing.

mikeserv

As of 3.11 they wrap their 64-bit cards in ... What's that thing for running 32-bit code on 64bit linux?

Multi something?

derobert

@mikeserv Where exactly are you getting that???

mikeserv

Hang on... Its harder on a phone.

Faheem Mitha

22:46

guys, i just renamed my xorg.conf file and will try rebooting without a file. is there anything else i need to do? i have xserver-xorg-video-nouveau installed

strugee

@mikeserv multilib

Faheem Mitha

actually, could i just restart the x server?

mikeserv

And it may have changed since they did it in... Oct or Nov or whenever. If so, that would be cool.

derobert

@FaheemMitha if you were previously using the nvidia proprietary drivers, you need to switch your glx back to mesa, and also change the module blacklist to not blacklist nouveau (and to blacklist nvidia instead)

Faheem Mitha

@derobert not sure what the former means.

what is glx and where is it set?

derobert

22:49

I think you use update-alternatives

but its changed depending on your Debian version...

on jessie, I think it's just update-alternatives --config glx

but I think there was a different way with wheezy

Faheem Mitha

@derobert also, not sure where to put the blacklisting. nouveau is blacklisted in nvidia-kernel-common.conf

i'm trying to find a guide. haven't found one yet

might be a reasonable question for the site, actually

derobert

indeed. I'd see if the update-alternative --config glx works

Faheem Mitha

@derobert shall i go ahead and post a question?

derobert

@FaheemMitha Sure. If Google hasn't easily given you an answer...

Faheem Mitha

@derobert nothing that is current, complete and accurate.

the debian wiki doesn't seem to have any thing up to date.

derobert

22:56

then even if it turns out to be as easy as that update-alternatives line, you can answer it it'll help the next person

Faheem Mitha

I wonder if this is relevant.

1

Q: Nouveau doesn't work after upgrade to Debian Wheezy

I upgraded from Debian squeeze to wheezy. My box contains an old graphics card not supported by current Nvidia drivers for distros newer than lenny (NV5M64 [RIVA TNT2 Model 64/Model 64 Pro]), so in my /etc/X11/xorg.conf, I had Driver "nv" specified in squeeze (which didn't seem to work after the ...

xorg kms nouveau

derobert

@FaheemMitha yeah, if you've disabled KMS, you'll need to turn it back on. Of course, if you were using the proprietary driver, it didn't have KMS, so you probably hand't disabled it.

Faheem Mitha

no, probably not

@derobert right, not an issue i imagine

@derobert ok, done

0

Q: Switching from Nvidia proprietary driver to Nouveau driver for Nvidia card

I have the following Nvidia graphics card installed: GeForce 7300 SE/7200 GS. I am currently using the proprietary Nvidia graphics drivers. I have the following packages installed containing nvidia in their names. ii glx-alternative-nvidia 0.2.2 amd64 ...

video nvidia nouveau

feel free to edit the question if you want

derobert

Looks good to me...

Faheem Mitha

@derobert If you can answer it, that would be helpful...

derobert

23:09

packages.debian.org/wheezy/glx-alternative-nvidia does say to use that update-alternatives line

that should also change the blacklist for you

Faheem Mitha

@derobert hmm.

root@orwell:/etc/modprobe.d# update-alternatives --config glx
There are 2 choices for the alternative glx (providing /usr/lib/glx).

Selection Path Priority Status
------------------------------------------------------------
* 0 /usr/lib/nvidia 100 auto mode
1 /usr/lib/mesa-diverted 5 manual mode
2 /usr/lib/nvidia 100 manual mode

derobert

you want to pick mesa-diverted

Faheem Mitha

@derobert ok, but why?

derobert

because MESA is the open source glx implementation, nvidia is the nvidia proprietary one

Faheem Mitha

@derobert ok. i'll try that now

derobert

23:11

its mesa-diverted because of the stunts the nvidia packages pull to install...

you'll need to change your xorg.conf... or probably just move it out of the way

Faheem Mitha

@derobert i still see `blacklist nouveau`
in nvidia-kernel-common.conf

@derobert i've moved that out of the way already

do i change the blacklist thing? if so, how?

derobert

ok, that must have changed... its in nvidia-blacklists-nouveau.conf in jessie

which goes away if you change to mesa

comment out the blacklist line, and run update-initramfs -u

Faheem Mitha

@derobert ok

derobert

to be paranoid, you could:

anthony@Zia:/etc/modprobe.d$ cat local-nouveau.conf.DISABLE
blacklist nvidia
blacklist nvidia_current

Faheem Mitha

@derobert what does update-initramfs -u do?

@derobert sounds reasonable

derobert

23:15

rebuilds the initramfs

of course, use without the .DISABLE extension... that's there because I switched back to the proprietary drivers once they fixed the signal mask bug

Faheem Mitha

@derobert and that is needed for what? because i changed the blacklist?

derobert

yes. The blacklist is copied to the initramfs. So you need to make sure the modules aren't loaded then

Faheem Mitha

@derobert ok

@derobert done. anything else?

oh, and do i need to reboot, or can i just restart X?

derobert

(I hear some people want video drivers early, which together with KMS allow them to display pretty pictures, instead of console boot message. They consider that a good thing, for some reason...)

@FaheemMitha I think if you stop X, rmmod nvidia, then you can probably modprobe nouveau and start X...

I have never tested that, though.

Faheem Mitha

@derobert maybe i'll just reboot

ok. shall i go ahead?

derobert

23:20

yeah, I think that's it. Your displays should at least come up... They may be in the wrong order. Or the wrong resolution... But they should work...

you can change all of that dynamically with xrandr

Faheem Mitha

@derobert ok, stand by

derobert

if you're left in terminal mode, copy the Xorg log somewhere so it doesn't get overwritten...

mikeserv

@derobert can't find it and boy woke. Park time. Best I can do is they're readme which discusses they're 4gb memory addressing limit. Maybe I'm mistaken, but it sure seems fishy that as soon as kernel enforces DMA they limit to 4gb.

derobert

/me hopes @FaheemMitha is OK, and hasn't died in a terrible reboot-induced incident.

derobert

23:46

You live!

Anyway, I'm off to get dinner....

If you ping me, it'll probably eventually get pushed over to my phone and I might notice it.

Transcript for