The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1028 »

Crusader: Apr 11, 2002

Has anyone ever had issues with the madwifi driver and the AR5212 chipset? I have a D-Link DWL-G510 that was pretty difficult to get stable with a Linksys dd-wrt AP:

Sep 25 14:34:19 box wpa_cli: interface ath0 DISCONNECTED
Sep 25 14:34:20 box dhcpcd[27270]: ath0: received SIGTERM, stopping
Sep 25 14:34:20 box dhcpcd[27270]: ath0: removing default route via 192.168.1.1 metric 2000
Sep 25 14:34:20 box dhcpcd[27270]: ath0: deleting IP address 192.168.1.102/24
Sep 25 14:34:20 box dhcpcd[27270]: ath0: exiting
Sep 25 14:34:23 box wpa_cli: interface ath0 CONNECTED
Sep 25 14:34:24 box dhcpcd[28193]: ath0: dhcpcd 3.0.16 starting
Sep 25 14:34:24 box dhcpcd[28193]: ath0: broadcasting for a lease
Sep 25 14:34:24 box dhcpcd[28193]: ath0: offered 192.168.1.102 from 192.168.1.1
Sep 25 14:34:24 box dhcpcd[28193]: ath0: leased 192.168.1.102 for 86400 seconds
Sep 25 14:34:24 box dhcpcd[28193]: ath0: adding IP address 192.168.1.102/24
Sep 25 14:34:24 box dhcpcd[28193]: ath0: adding route to 192.168.1.0 (255.255.255.0) via 0.0.0.0 metric 2000
Sep 25 14:34:24 box dhcpcd[28193]: ath0: removing route to 192.168.1.0 (255.255.255.0) via 0.0.0.0 metric 0
Sep 25 14:34:24 box dhcpcd[28193]: ath0: adding default route via 192.168.1.1 metric 2000

Over and over, at random. At first I tried switching channels and positioning the antenna to raise the overall signal, but it would still drop out even with a -47dBm to -95dBm ratio.

Eventually, doing all this got it to where I wouldn't have packet loss just pinging the router, and the random dropouts ceased:

Disabling protected mode, bursting, fast frames, and background scanning on the card, and lowering the ACK timing (to 0) and beacon interval (to 75ms) on the router.

I was mainly wondering if anyone else had a similar experience. Hopefully the ath5k driver becomes a viable replacement relatively shortly (I'm pretty certain it's the card driver since my other wireless devices didn't have this trouble).

Crusader fucked around with this message at 06:24 on Sep 26, 2007

# ? Sep 26, 2007 06:21

Adbot: ADBOT LOVES YOU

# ? May 10, 2024 02:45

deimos: Nov 30, 2006; Forget it man this bat is whack, it's got poobrain!

No problems here:

code:

05:01.0 Ethernet controller: Atheros Communications, Inc. AR5212 802.11abg NIC (rev 01)

It's pretty drat stable all things considered (my house is a loving interference nightmare). I am not sure if I have a DWL 510 or 520.

Did it work before switching to Linux? Maybe it's a dd-wrt power setting. Try flipping the antennas from the router to your adapter.

# ? Sep 26, 2007 14:54

Crusader: Apr 11, 2002

deimos posted:

Did it work before switching to Linux? Maybe it's a dd-wrt power setting. Try flipping the antennas from the router to your adapter.

It's only ever been a Linux machine; I did try bumping the transmission power, but the router isn't very far from everything else and I was seeing overall signal degradation compared to the default 28mW setting.

I had a homemade parabolic reflector that I tried on both the router and the adapter; the SNR was higher, but I was still getting the inexplicable dropouts :/

Either way, it's alright now with the setting tweaks mentioned above - I just wish I had a better explanation for what the issue was in the first place

# ? Sep 26, 2007 15:57

Prince John: Jun 20, 2006; Oh, poppycock! Female bandits?

teapot posted:

Looks pretty normal for me, so defaults should be usable.

In addition to removing or renaming .gnome , .gnome2 and .gnome2_private you need to kill gconfd-2 process, then remove or rename .gconf and .gconfd directories. On the next login they will be re-created with defaults.

How very bizarre. It reset most of my settings, but it didn't get rid of everything (the Home icon on the desktop was using an old theme for example) and the problem with the Places menu still existed.

I created a different user account as a test and everything worked perfectly. Even more surprising is that my mysterious unfixable HAL and gnome-volume-manager problem that I did a thread about in the Arch forums is also fixed in the new account.

Obviously my profile has just gone a bit funny, so think I will just shift my home folder and start fresh. Thanks for your help!

Prince John fucked around with this message at 18:57 on Sep 26, 2007

# ? Sep 26, 2007 18:52

teapot: Dec 27, 2003; by Fistgrrl

Col posted:

How very bizarre. It reset most of my settings, but it didn't get rid of everything (the Home icon on the desktop was using an old theme for example) and the problem with the Places menu still existed.

I created a different user account as a test and everything worked perfectly. Even more surprising is that my mysterious unfixable HAL and gnome-volume-manager problem that I did a thread about in the Arch forums is also fixed in the new account.

Obviously my profile has just gone a bit funny, so think I will just shift my home folder and start fresh. Thanks for your help!

Sometimes this happens if gconfd-2 or some other piece of software remains running after logout and re-creates the old defaults.

Also I think, I have missed .local in my lists.

# ? Sep 26, 2007 19:11

Cabbages and VHS: Aug 25, 2004; Listen, I've been around a bit, you know, and I thought I'd seen some creepy things go on in the movie business, but I really have to say this is the most disgusting thing that's ever happened to me.

can anyone point me to an x display manager that is very basic, runs command line and just does one application fullscreen at a time (that is, exit/crash back to shell)? Is there an easy way to do such a thing?

# ? Sep 27, 2007 02:27

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

Quick question: Is there an easy way to send an AOL IM from a script? I'm looking for something like "echo 'Hewo new yorku!' | sendim -u username -p password -d 1337Hiro". The closest I've found is a bare-bones stateful script, which requires me to log in, send the IM and then log out (though, come to think of it, I don't know if it even provides a log out command.)

# ? Sep 27, 2007 02:42

GringoGrande: Jul 27, 2001; Nah...

Magicmat posted:

Quick question: Is there an easy way to send an AOL IM from a script? I'm looking for something like "echo 'Hewo new yorku!' | sendim -u username -p password -d 1337Hiro". The closest I've found is a bare-bones stateful script, which requires me to log in, send the IM and then log out (though, come to think of it, I don't know if it even provides a log out command.)

You should be able to do that with pidgin via it's dbus interface. It's kind of under documented, but it shouldn't be that hard to write up something in python or a shell script using purple-send.

edit: purple-remote seems to be what you are looking for.

code:

#!/bin/bash
#send-im <name> <message>
purple-send 'icq:goim?screenname=$1&message=$2

GringoGrande fucked around with this message at 03:43 on Sep 27, 2007

# ? Sep 27, 2007 03:24

teapot: Dec 27, 2003; by Fistgrrl

PFC.Spengler posted:

can anyone point me to an x display manager that is very basic, runs command line and just does one application fullscreen at a time (that is, exit/crash back to shell)? Is there an easy way to do such a thing?

startx and $HOME/.xinitrc

You will likely have to give the application geometry arguments on the command line, and if it uses any kind of additional windows the interface is likely to suck. Fortunately you can run sawfish as your window manager and tell it that your application's main window is supposed to be always fullscreen with no decorations while all other windows are managed like they usually are.

# ? Sep 27, 2007 03:43

Jonny 290: May 5, 2005; [ASK] me about OS/2 Warp

I've got a Feisty Desktop (but server-ized and streamlined) install with some NTFS drives for my stashed media (Pulled out of my old XP server). I'm getting silly "operation not permitted" errors when moving or altering files, evidently because it's not happy with preserving timestamps. I've done the setuid root shuffle, to no avail. It's just annoying.

Would I be better off nuking and going FAT32 with these drives? I want to maintain cross-compatibility.

Edit: Also, I want to migrate my install to a 2x 9gb scsi drive setup (/ and /home). I installed using the boring, vanilla Ubuntu CD. What sort of hoops am I going to have to jump through to get it sorted out? I figure format the drives, install grub (how?) and copy the partitions? Better done via a liveCD boot, perhaps?

# ? Sep 27, 2007 03:53

teapot: Dec 27, 2003; by Fistgrrl

Jonny 290 posted:

I've got a Feisty Desktop (but server-ized and streamlined) install with some NTFS drives for my stashed media (Pulled out of my old XP server). I'm getting silly "operation not permitted" errors when moving or altering files, evidently because it's not happy with preserving timestamps. I've done the setuid root shuffle, to no avail. It's just annoying.

Actually this is because you use the original kernel implementation of NTFS instead of ntfs-3g.

quote:

Would I be better off nuking and going FAT32 with these drives? I want to maintain cross-compatibility.

Edit: Also, I want to migrate my install to a 2x 9gb scsi drive setup (/ and /home). I installed using the boring, vanilla Ubuntu CD. What sort of hoops am I going to have to jump through to get it sorted out? I figure format the drives, install grub (how?) and copy the partitions? Better done via a liveCD boot, perhaps?

You can do it from running system, however it would be very easy from a live CD:

Boot from CD when all drives are attached.

Once you have formatted the drives (in this example source is /dev/sda1, target filesystems are on /dev/sdb1 and /dev/sdc1), run "sudo -s" on console or in terminal to become root. Mount all filesystems and copy installation using tar, so it will preserve all permissions:

code:

mkdir /mnt/source
mount /dev/sda1 /mnt/source
mkdir /mnt/target
mount /dev/sdb1 /mnt/target
mkdir /mnt/target/home
mount /dev/sdc1 /mnt/target/home
cd mnt/target
(cd /mnt/source; tar c . )| tar xv
cd /

Edit new fstab:

code:

vi /mnt/target/etc/fstab

You should either use device names how they will appear when only new disks will be installed, or use UUIDs that you can find by running tune2fs on your new filesystems.

code:

umount /mnt/target/home
umount /mnt/target
umount /mnt/source

(you can omit "v" in tar command if you don't want tar to waste time displaying the list of files).

Shut down, move all drives to their final positions, boot again with a live CD. In root shell mount the drives again (their device names should be changed to what they will become without the original drive) and mount /proc:

code:

mkdir /mnt/target
mount /dev/sda1 /mnt/target
mkdir /mnt/target/home
mount /dev/sdb1 /mnt/target/home
mount --bind /proc /mnt/target/proc

Then chroot into the new environmrnt:

code:

chroot /mnt/target bash

At that point you can edit grub configuration files and install grub on the new device using your own system running in chroot.

code:

grub-install /dev/sda

exit from this environment, unmount filesystems:

code:

exit
umount /mnt/target/proc
umount /mnt/target/home
umount /mnt/target

Reboot the system without a live CD. If any problems happen at that point, you can repeat the last step to be able to edit files on your system while running it in a chroot environment.

# ? Sep 27, 2007 04:25

Jonny 290: May 5, 2005; [ASK] me about OS/2 Warp

Thank you, you rule so much.

My only concern is going to be the /dev/ names, as my disks are currently ID'd sda/sdb. I'll knock this out tonight, though!

# ? Sep 27, 2007 04:33

JoeNotCharles: Mar 3, 2005; Yet beyond each tree there are only more trees.

Every once in a while at work I end up with massive cache or log directories full of millions of tiny files. "rm -rf *" takes forever to get rid of these - is there any tool that'll speed this up?

(I just realized if I rename the directory to something else, I can at least move it out of the way and then delete it in the background while I do other tasks, but I'd still like to free up space a bit faster.)

# ? Sep 27, 2007 16:57

Swordfish: Aug 15, 2001

Since there doesn't seem to be a FreeBSD questions thread, a quick one:

We have a machine with a new 3ware 9550SXU RAID controller. FreeBSD 6.2 goes into an infinite reboot loop whenever it loads the driver. Apparently for everyone else on the internet, this doesn't happen. Any ideas?

# ? Sep 27, 2007 20:31

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

GringoGrande posted:

You should be able to do that with pidgin via it's dbus interface. It's kind of under documented, but it shouldn't be that hard to write up something in python or a shell script using purple-send.

edit: purple-remote seems to be what you are looking for.
code:
#!/bin/bash
#send-im <name> <message>
purple-send 'icq:goim?screenname=$1&message=$2

Am I correct in assuming that that would require Pidgin to be running to work? I really need something that exists entirerly in a single command, with no background processes or additional commands to execute.

# ? Sep 28, 2007 03:40

Scaevolus: Apr 16, 2007

Magicmat posted:

Am I correct in assuming that that would require Pidgin to be running to work? I really need something that exists entirerly in a single command, with no background processes or additional commands to execute.

AIM isn't really very well suited for one-shot messages.

# ? Sep 28, 2007 03:42

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

Scaevolus posted:

AIM isn't really very well suited for one-shot messages.

Yeah, I know. But it is the most convenient for me. If there's not an easy way to do it via script, maybe I'll just fire off an e-mail to a e-mail-to-AOL gateway out there. Or maybe if I get up the motivation, I'll hack something up myself.

# ? Sep 28, 2007 03:58

teapot: Dec 27, 2003; by Fistgrrl

JoeNotCharles posted:

Every once in a while at work I end up with massive cache or log directories full of millions of tiny files. "rm -rf *" takes forever to get rid of these - is there any tool that'll speed this up?

(I just realized if I rename the directory to something else, I can at least move it out of the way and then delete it in the background while I do other tasks, but I'd still like to free up space a bit faster.)

You can put rm into background as

code:

rm -rf *&

"*" is resolved by the shell when it encounters command line and before it actually runs rm, so if any new files will be created in this directory while rm is running, they won't be deleted. The time that rm takes to remove files is entirely based on filesystem and device performance -- space previously used by files should be marked as free, so it will be reused for new files. While in theory you can speed it up by trying to remove multiple files in parallel as opposed to doing it in a sequence, it is unlikely to get noticeably faster.

# ? Sep 28, 2007 04:11

teapot: Dec 27, 2003; by Fistgrrl

Magicmat posted:

Yeah, I know. But it is the most convenient for me. If there's not an easy way to do it via script, maybe I'll just fire off an e-mail to a e-mail-to-AOL gateway out there. Or maybe if I get up the motivation, I'll hack something up myself.

Email is also dependent on a process (your mail server) running somewhere, so I don't see an advantage.

# ? Sep 28, 2007 04:14

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

teapot posted:

Email is also dependent on a process (your mail server) running somewhere, so I don't see an advantage.

No, not my mail server, my ISP's. I can just use 'mail' to fire off an e-mail using my ISP's SMTP server, rather than running a SMTP server myself, which is what I already use for other notifications. Maybe not Enterprise Quality[tm], but for my personal router, it works.

# ? Sep 28, 2007 04:17

teapot: Dec 27, 2003; by Fistgrrl

Swordfish posted:

Since there doesn't seem to be a FreeBSD questions thread, a quick one:

We have a machine with a new 3ware 9550SXU RAID controller. FreeBSD 6.2 goes into an infinite reboot loop whenever it loads the driver. Apparently for everyone else on the internet, this doesn't happen. Any ideas?

Update the driver to the latest version -- there were updates since 6.2-RELEASE:

http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/106488

# ? Sep 28, 2007 04:37

Xenomorph: Jun 13, 2001

I used G4L / Ghost 4 Linux today, and Hard Drive performance from a 5400RPM ATA33 IDE drive was about 17 megabytes/second for reading.

Under Windows, this drive benched 30 megabytes/sec sustained over USB2.

Does anyone know if G4L has performance issues? I tried direct to disk copying, and from the disk to the network (just to make sure it wasn't an issue with the system writing to the other drive). It would not get past 17 megabytes/sec.

# ? Sep 28, 2007 05:28

teapot: Dec 27, 2003; by Fistgrrl

Xenomorph posted:

I used G4L / Ghost 4 Linux today, and Hard Drive performance from a 5400RPM ATA33 IDE drive was about 17 megabytes/second for reading.

Under Windows, this drive benched 30 megabytes/sec sustained over USB2.

Does anyone know if G4L has performance issues? I tried direct to disk copying, and from the disk to the network (just to make sure it wasn't an issue with the system writing to the other drive). It would not get past 17 megabytes/sec.

Use hdparm <device> to see the drive configuratuib and hdparm -Tt <device> to check the actual hard drive performance. It's possible that you have DMA or other modes supported by the drive turned off.

# ? Sep 28, 2007 05:55

deimos: Nov 30, 2006; Forget it man this bat is whack, it's got poobrain!

Xenomorph posted:

I used G4L / Ghost 4 Linux today, and Hard Drive performance from a 5400RPM ATA33 IDE drive was about 17 megabytes/second for reading.

Under Windows, this drive benched 30 megabytes/sec sustained over USB2.

Does anyone know if G4L has performance issues? I tried direct to disk copying, and from the disk to the network (just to make sure it wasn't an issue with the system writing to the other drive). It would not get past 17 megabytes/sec.

did you use both through USB2 or did you plug it in to IDE on Linux? Maybe you used a 40 cable ribbon instead of an 80?

# ? Sep 28, 2007 06:01

fischtick: Jul 9, 2001; CORGO, THE DESTROYER; Fun Shoe

Magicmat posted:

Quick question: Is there an easy way to send an AOL IM from a script? I'm looking for something like "echo 'Hewo new yorku!' | sendim -u username -p password -d 1337Hiro". The closest I've found is a bare-bones stateful script, which requires me to log in, send the IM and then log out (though, come to think of it, I don't know if it even provides a log out command.)

Do you do perl? There's the Net::AIM module that will let you create exactly what you're describing, using AOL's TOC network protocol. It's a fairly simple module to work with.

That said, I haven't touched perl or Net::AIM in like 3 years. It might not work anymore.

# ? Sep 28, 2007 07:42

Steve French: Sep 8, 2003

Magicmat posted:

Am I correct in assuming that that would require Pidgin to be running to work? I really need something that exists entirerly in a single command, with no background processes or additional commands to execute.

You might be able to do the same thing using finch; it wouldn't all be one command, but if its in a shell script, you could just start finch, send the message, close finch.

Having never actually used finch (just pidgin) I'm just kinda stabbing in the dark here, but it might be something to look into.

# ? Sep 28, 2007 08:22

stubblyhead: Sep 13, 2007; That is treason, Johnny!; Fun Shoe

I am running Feisty ubuntu on my Dell Inspiron laptop, and am very pleased for the most part. I am having some trouble with my networking, however (I primarily use wifi). After I start up, everything works fine, but after maybe 10 minutes or so, my web connectivity goes away. Page loads time out, but my network connection icon still shows that I am connected. I think it only affects some ports, because telnet and irc don't seem to be affected. To get my networking going again though, I have to disable and enable it again, and it works fine indefinitely.
Another problem that may or may not be related is I often cannot connect to my wireless router after restoring from hibernation. I get the network icon with the error mark on it, and when I try to manually reestablish connection, it seems like it tries for a couple seconds and then quits. Sometimes my wifi is just fine, though. Any suggestions on what I can do to fix this or what I should do to better troubleshoot it?

# ? Sep 28, 2007 08:59

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

OK, I hacked up a simple .NET app that runs fine under Mono to send a single IM. It works great except for one thing: I've set $MONO_PATH in /etc/bash/bashrc (Gentoo system) and in /etc/profile (side note: What is the proper way to set an environmental variable for all users? My shell-fu is weak) so that sudo will have the correct $MONO_PATH. And executing "sudo echo $MONO_PATH" does, indeed, produce the desired results. And putting "echo $MONO_PATH" in a bash script also produces the desired results. But putting the command in a bash script and then executing that script as sudo gives me an empty string. What gives? How can I have my bash scripts executed with sudo retain that environmental variable?

# ? Sep 28, 2007 12:29

Xenomorph: Jun 13, 2001

teapot posted:

Use hdparm <device> to see the drive configuratuib and hdparm -Tt <device> to check the actual hard drive performance. It's possible that you have DMA or other modes supported by the drive turned off.

I'll see about trying this today.

Edit: I just did hdparm on the drive. It said this:
Timing buffer-cache reads: 7188 MB in 3.00 seconds = 2397.02 MB/sec
Timing buffered disk reads: 96 MB in 3.00 seconds = 31.96 MB/sec

That 31 megs/second thing looks more like what I was expecting.
However, I'm doing another run with g4l, and HD reading has again only made it up to about 19 megabytes/sec. It's connected via USB to the Linux system now, and is doing better than the 17megs/sec it was doing when connected via IDE yesterday.

I ran "hdparm" from the G4L boot CD I have, so it's not like the configuration is suddenly changing on it.

deimos posted:

did you use both through USB2 or did you plug it in to IDE on Linux? Maybe you used a 40 cable ribbon instead of an 80?

On the system running Linux, it was installed with an 80 wire IDE cable to the system's internal primary IDE port.
Wouldn't a 40 wire cable still support 33megs?

Xenomorph fucked around with this message at 13:40 on Sep 28, 2007

# ? Sep 28, 2007 12:52

teapot: Dec 27, 2003; by Fistgrrl

Magicmat posted:

OK, I hacked up a simple .NET app that runs fine under Mono to send a single IM. It works great except for one thing: I've set $MONO_PATH in /etc/bash/bashrc (Gentoo system) and in /etc/profile (side note: What is the proper way to set an environmental variable for all users? My shell-fu is weak) so that sudo will have the correct $MONO_PATH. And executing "sudo echo $MONO_PATH" does, indeed, produce the desired results. And putting "echo $MONO_PATH" in a bash script also produces the desired results. But putting the command in a bash script and then executing that script as sudo gives me an empty string. What gives? How can I have my bash scripts executed with sudo retain that environmental variable?

1. sudo sanitizes the environment, and not doing so would be a huge security hole.
2. When it runs shell, shell re-creates the environment by running /etc/profile .
3. "sudo echo $MONO_PATH", being a shell command, is converted to "sudo echo /my/actual/path/to/mono" by the non-privileged shell that runs it.
4. You are not supposed to run anything but administrative utilities by sudo, and certainly not IM scripts.
5. Put your variables assignment in the same shell script that runs everything else. If you have to use sudo, run that script with sudo, however if it accepts arguments most likely it is a security hole.

# ? Sep 28, 2007 13:09

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

teapot posted:

1. sudo sanitizes the environment, and not doing so would be a huge security hole.
2. When it runs shell, shell re-creates the environment by running /etc/profile .
3. "sudo echo $MONO_PATH", being a shell command, is converted to "sudo echo /my/actual/path/to/mono" by the non-privileged shell that runs it.
4. You are not supposed to run anything but administrative utilities by sudo, and certainly not IM scripts.
5. Put your variables assignment in the same shell script that runs everything else. If you have to use sudo, run that script with sudo, however if it accepts arguments most likely it is a security hole.

So the answer is to assign the variable in the script? Got it.

And the IM app certainly is run by an administrative script. Some people have admin scripts which e-mail them when they're done; mine IMs me. Stop making stupid assumptions.

# ? Sep 28, 2007 13:15

teapot: Dec 27, 2003; by Fistgrrl

Magicmat posted:

So the answer is to assign the variable in the script? Got it.

And the IM app certainly is run by an administrative script. Some people have admin scripts which e-mail them when they're done; mine IMs me. Stop making stupid assumptions.

When you run something as complex and potentially insecure as mono from a script running as root, you are supposed to su to a non-privileged user from your script, having another script with insecure stuff as su argument. That second script will include your environment variables assignment and calls to IM-related utilities.

# ? Sep 28, 2007 13:23

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

teapot posted:

When you run something as complex and potentially insecure as mono from a script running as root, you are supposed to su to a non-privileged user from your script, having another script with insecure stuff as su argument. That second script will include your environment variables assignment and calls to IM-related utilities.

Hmm, that's actually a pretty good idea. I'll do that, thanks!

Edit: If I were to need to do something like this again, but using only a single command instead of the two I need here, would sudo -u user be acceptable instead of su user? And, inside a script, is the "exit" command the recommended way to exit the su context?

Magicmat fucked around with this message at 13:38 on Sep 28, 2007

# ? Sep 28, 2007 13:30

teapot: Dec 27, 2003; by Fistgrrl

Magicmat posted:

Edit: If I were to need to do something like this again, but using only a single command instead of the two I need here, would sudo -u user be acceptable instead of su user?

Only if for some reason you are not root when issuing it (different users with different privileges but none of them is root).

It's possible to create a sudoers file that will not allow root to sudo as any other user, however su always works for root.

quote:

And, inside a script, is the "exit" command the recommended way to exit the su context?

When su runs a script, the whole script run as the user given to su, exit terminates the script in the same way as reaching the end of it. Your original script that called su remains running as root, waiting for su to complete, so after su it will continue as root. Same applies to original user that called sudo -- script that sudo runs is running as root, however when sudo exits, original shell is at the same user that called sudo.

I also forgot to mention that when sudo sanitizes the environment it does not remove all variables, and MONO_PATH is not among ones that it clears (though likely it will be added to that list as security-sensitive, along with PERLLIB and JAVA_TOOL_OPTIONS that serve similar purpose). At least for now you can use

code:

export MONO_PATH

to pass it to sudo.

# ? Sep 28, 2007 14:34

Magicmat: Aug 14, 2000; I've got the worst fucking attorneys

teapot posted:

Only if for some reason you are not root when issuing it (different users with different privileges but none of them is root).

It's possible to create a sudoers file that will not allow root to sudo as any other user, however su always works for root.

Why can't I be root? Would that just be bad security practice for some reason? Practically, what is the difference between sudo -u user ./script.sh and su user -c ./script.sh?

quote:

I also forgot to mention that when sudo sanitizes the environment it does not remove all variables, and MONO_PATH is not among ones that it clears (though likely it will be added to that list as security-sensitive, along with PERLLIB and JAVA_TOOL_OPTIONS that serve similar purpose). At least for now you can use
code:
export MONO_PATH
to pass it to sudo.

Wait, I'm not sure I'm understanding you. Do you mean that my original example should have worked? That doing sudo ./script.sh, where script.sh contains "echo $MONO_PATH" should have the correct output? Or do you mean something else?

# ? Sep 28, 2007 15:20

teapot: Dec 27, 2003; by Fistgrrl

Magicmat posted:

Why can't I be root? Would that just be bad security practice for some reason? Practically, what is the difference between sudo -u user ./script.sh and su user -c ./script.sh?

Usually /etc/sudoers file contains the line

code:

root    ALL=(ALL) ALL

If that line is removed, sudo will fail when root runs sudo -u user
On the other hand, su always works for root.

quote:

Wait, I'm not sure I'm understanding you. Do you mean that my original example should have worked? That doing sudo ./script.sh, where script.sh contains "echo $MONO_PATH" should have the correct output? Or do you mean something else?

It would work if you ran

code:

MONO_PATH=/path/to/mono
export MONO_PATH
sudo ./script.sh

however it's possible that it will stop working if at some point sudo authors decided that MONO_PATH shouldn't be passed because it is at least as security-sensitive as other variables that it cleans now.

teapot fucked around with this message at 00:25 on Sep 29, 2007

# ? Sep 28, 2007 16:04

Kobayashi: Aug 13, 2004; by Nyc_Tattoo

The amount of knowledge contained in this thread is amazing.

I have two questions:

1. VirtualBox keeps killing my VM because the image has grown to be 16 Gb. How can I increase the maximum filesize for my system? I run stock Ubuntu Feisty, shipped to me by Dell. I believe that means ext3.

2. I cannot reboot. When I try to reboot, the system hangs after stopping all the running processes. At first I couldn't see what was happening because of that stupid graphical progress bar that obscures what is actually happening in the terminal. I then booted into the recovery console and tried to reboot. The system just sits there after saying "The system will now restart." This is particularly annoying since I do not have a reset button, so I have to hold power for ~7 seconds to turn off my machine. Could this be an ACPI issue? Halt works fine.

# ? Sep 28, 2007 21:19

Steve French: Sep 8, 2003

Can anybody tell me why the hell this is happening and/or how to fix it?

(Yes, the drive is actually 465.8GB)

EDIT: Nevermind, figured it out, there was somehow 250 gb of poo poo in the trash folder.

Steve French fucked around with this message at 08:45 on Sep 29, 2007

# ? Sep 29, 2007 08:40

Crush: Jan 18, 2004; jot bought me this account, I now have to suck him off.

Steve French posted:

Can anybody tell me why the hell this is happening and/or how to fix it?

(Yes, the drive is actually 465.8GB)

EDIT: Nevermind, figured it out, there was somehow 250 gb of poo poo in the trash folder.

Pacman!

# ? Sep 29, 2007 09:13

Adbot: ADBOT LOVES YOU

# ? May 10, 2024 02:45

teapot: Dec 27, 2003; by Fistgrrl

Kobayashi posted:

The amount of knowledge contained in this thread is amazing.

I have two questions:

1. VirtualBox keeps killing my VM because the image has grown to be 16 Gb. How can I increase the maximum filesize for my system? I run stock Ubuntu Feisty, shipped to me by Dell. I believe that means ext3.

Can you clarify, which OS runs VirtualBox, and which OS is running in the VirtualBox? And where do you keep their corresponding filesystems?

quote:

2. I cannot reboot. When I try to reboot, the system hangs after stopping all the running processes. At first I couldn't see what was happening because of that stupid graphical progress bar that obscures what is actually happening in the terminal. I then booted into the recovery console and tried to reboot. The system just sits there after saying "The system will now restart." This is particularly annoying since I do not have a reset button, so I have to hold power for ~7 seconds to turn off my machine. Could this be an ACPI issue? Halt works fine.

Very likely it's ACPI, however it's possible that some driver does not exit cleanly (then you most likely will see error messages about it while computer is waiting and is supposed to be rebooting). You can press 'e' in GRUB menu while your normal system is selected, select the kernel line, pres 'e' again, edit the line to remove "splash" and "quiet", press Enter and 'b' to boot in normal mode with kernel messages enabled. You can do the same way to add "acpi=noirq" or even "acpi=off" options to see if it makes reboot work properly.

# ? Sep 29, 2007 10:27

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Linux Questions Thread: a bunch of pitfalls, but technically it's possible

«‹›1028 »