'cp' keeps getting stuck...

Message

Utoxin · Post by **Utoxin** » Wed Jun 12, 2002 9:30 pm

Several times in the last couple days, when copying things as root, the cp process has hung, and the process refuses to die, even when killed with a -9. Eventually, the process dies on it's own, but it takes forever.

My filesystem is XFS, and I'm running with gcc3.1. Anyone have any ideas?

Post by **pjp** » Wed Jun 12, 2002 9:32 pm

What are you trying to copy... how large are the files?

lx · Post by lx » Wed Jun 12, 2002 9:40 pm

Well that's one of the reasons I stopped using XFS. Used it on some partitions, and I found out that copying a large file (670Mb), from one partition to another hung the system/cp. Well didn't really fix it, just dumped it for ReiserFS.

Good Luck, lX

Utoxin · Post by **Utoxin** » Wed Jun 12, 2002 9:43 pm

Well, the one today is only 21 megs, and the copy process has used up half an hour of processor time so far. Total, it's been running for an hour.

lx · Post by lx » Wed Jun 12, 2002 10:00 pm

Utoxin wrote:Well, the one today is only 21 megs, and the copy process has used up half an hour of processor time so far. Total, it's been running for an hour.

Well I wasn't that patient (xfs was probably in a loop) I just did CTRL+C and did another cp and it worked. I concluded to dump Xfs.

Cya lX.

Utoxin · Post by **Utoxin** » Wed Jun 12, 2002 10:05 pm

I'd do that, but I've got too much time invested in this system to rebuild it.

Anyone have any ideas why this happens, or how to fix it?

Post by **klieber** » Wed Jun 12, 2002 10:52 pm

You say it hangs when copying stuff as root. Does it work fine when copying stuff as a normal user?

Also, I seriously doubt this is the problem, but make sure you have DMA enabled on the device. (hdparm -d /dev/<your partition>)

--kurt

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 12:04 am

klieber wrote:You say it hangs when copying stuff as root. Does it work fine when copying stuff as a normal user?

Also, I seriously doubt this is the problem, but make sure you have DMA enabled on the device. (hdparm -d /dev/<your partition>)

--kurt

So far, I haven't seen it hang while running as a normal user.

And yes, DMA is enabled.

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 12:54 am

Well, I eventually got the thread to die, and started the gcc emerge again (It's what was running when it locked up last.) And lo-and-behold, it locked up on exactly the same file copy the second time. This to me sounds like a bug of somekind in XFS itself. Where would I go to report XFS bugs?

Post by **pjp** » Thu Jun 13, 2002 1:47 am

Is it just this one file? Can you try another file of similar size and see if it locks up?
May sound weird, but I had a file one time that crashed a Novell file server whenever
I tried to open it. At the time, we didn't know this, but it was eventually discovered
that it was this one file. After recreating the file with essentially the same contents,
the problem never reoccurred.

jtanner · Post by **jtanner** » Thu Jun 13, 2002 1:53 am

Just to make sure: Neither the source nor destination is on a network mounted filesystem (nfs, samba, etc) or on removable media, right?

Jim

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 2:01 am

I let it run this time, and it eventually made it through that file. And one of the next files was twice that size, and it's hanging there too. So it's something to do with large files. (Where large seems to mean >~20MB)

And no, it's not over the network or anything. It's all on the same hard disk.

jtanner · Post by **jtanner** » Thu Jun 13, 2002 2:28 am

Darn, I thought this was going to be easy!

How full is your disk? Any issues/info with XFS and GCC 3.1?

Well, when kill -9 fails to destroy a process, it's because it's blocking in the kernel (duh). Could this be a tunable parameter in XFS? I've never used it, but I seem to remember reading that there are a bunch of tunable parameters that can make a big difference performance-wise.

Isn't XFS supposed to be GOOD with large files?

Jim

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 2:33 am

Disk has gigs and gigs of space left.

There's no specific issues with GCC3.1, but they did say they didn't gauruntee it to work. I'm beginning to suspect that's at the root of the problem.

And there's not much on XFS in the kernel config. Mostly quota stuff. Which I have turned off.

Post by **pjp** » Thu Jun 13, 2002 2:36 am

I recall an fstab setting we used for Solaris that was related to large files, but I don't recall what it was. Any one know if this could be similar?

jtanner · Post by **jtanner** » Thu Jun 13, 2002 2:42 am

I know this sucks, but....maybe try to recompile with GCC 2.95.3, and see what happens. You would just have to recompile the kernel and the tools, right?

Jim

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 2:45 am

I'd also have to compile gcc 2.95. Wanna guess how long that'd take in my system's current condition?

jtanner · Post by **jtanner** » Thu Jun 13, 2002 2:51 am

Ummm....no.

Precompiled binaries, maybe? Besides, it beats having to reformat your drive...

If you have a spare disk sitting around, maybe set that up as ext2/3 and use it to build the compiler, at least.

Jim

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 9:26 pm

Well, I finally finished the upgrade to the latest gcc3, and did an mrproper compile of my kernel, and things seem to have improved. I'll let everyone know if I have any more problems.

lx · Post by lx » Thu Jun 13, 2002 9:49 pm

I remember having the same problems with Xfs (as I stated in the earlier post) also I got error with the preempt patch + xfs. Some pre-empt count wasn't zero after finishing some command (cp / something else), so I just dumped it, that's the kind of guy I am.....

. Well I want a stable and reliable filesystem, not even one cp instruction may hang..... Are you getting the pre-empt count error.

Cya lX.

Utoxin · Post by **Utoxin** » Thu Jun 13, 2002 10:12 pm

Preempt count is /rarely/ 0. It's not an error. It just means the kernel let something else take over the processor while it was running.

mark · Post by **mark** » Sat Jun 15, 2002 2:14 pm

Its probably not related but you never know. When I had ACL in my use variables and compiled fileutils I would subsequently get error messages with cp -p. This has been reported to xfs who are putting some changes in cvs. They are also liasing with fileutils maintainers and the maintainers of a gentoo fileutils patch to try to find the route cause.

If your not using acl's you could see if use=ACL is set and disable it if it is. Then re emerge fileutils. I am not running gcc3 and my symptoms werent the same but it could be worth a try.

Mark

'cp' keeps getting stuck...

'cp' keeps getting stuck...

I had probs with cp -p and xfs