Thought I was past the stage of posting such basic questions but obviously I am not...

I have started to get "No space left on device" errors when writing to files in /home (/dev/hda9 on my box). But df says there is plenty of space and lots of free i-nodes:

$ cp yyy zzz
cp: writing `zzz': No space left on device
$ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/hda7 4.9G 1.9G 2.8G 40% /
none 125M 0 124M 0% /dev/shm
/dev/hda9 21G 3.2G 16G 17% /home
/dev/hda1 5.9G 4.6G 1.3G 78% /mnt/windows/system
/dev/hda5 2.9G 2.2G 830M 73% /mnt/windows/d
/dev/hda6 1.2G 874M 348M 72% /mnt/windows/e
$ df -i
Filesystem Inodes IUsed IFree IUse% Mounted on
/dev/hda7 656000 147226 508774 23% /
none 31940 1 31939 1% /dev/shm
/dev/hda9 5672960 18813 5654147 1% /home
/dev/hda1 0 0 0 - /mnt/windows/system
/dev/hda5 0 0 0 - /mnt/windows/d
/dev/hda6 0 0 0 - /mnt/windows/e

What might be going on here?

may you've limited the free space size for the regular users via quota,can you copy to that partition as a root or not?

Try df -i
perhaps you're out of inodes.
(as I recall it will report as "No space left on device")

I did think it might be quotas but the quota command reports "none" for all users (both -u and -g so it's not group quotas either).

I-nodes are not the problem.
Any other ideas?

Thought I'd post the resolution...

It turned out to be a corrupted file system. Calling e2fsck from single used mode fixed it. I spotted that when the system seemed out of space, a kernel error messge from the file system module reported that there was an ambiguity in something called a block group.
I didn't think of running e2fsck because it was run when the machine rebotted. Obviously it get run in such a way that not all errors get fixed.

I have same kind of issue,

I have ran e2fsck -fyc <device> (/dev/emcpowera1)

than mount the partition, and do #cd lost+found, there u will find some recovered data.

i my case these data was not required thus we have remove the data and recover the space.