This is exactly the same bug as #30944, which was reported more than a month ago but still is unfixed in 7.1. I am raising its severity and priority to high and resubmitting it as a RedHat 7.1 bug. One additional comment: to see this bug, you should do "ls *" or "echo *" in tcsh, or use the glob() function in perl; doing "ls *" and "echo *" in bash will not reveal the bug as bash seems to have its own way of doing file glob'ing.
Created attachment 16346 [details] Test program that reproduces bug
We have the same problem on our site. I've just attached a test program that reproduces it. Also, http://www.fys.uio.no/~trondmy/src/2.4.3/ contains code that fixes the problem.
I've patched the rh71 kernel rpms and made some new iso files with these. They are available at ftp://ftp.crc.dk/pub/rh71irixnfspatch Note: They are without any warranty whatsoever, use them at your own risk!
The patch should also be in kernel version 2.4.3-2.12.1 and later which is in rawhide / should be there soon
I actually tried downloading and installing kernel-2.4.3-2.12.1 from rawhide, but I still had the same problem. Perhaps the patch was introduced after 2.4.3-2.12.1 was released?
2.4.3-5 in rawhide absolutely has the patch. If it still doesn't work it's a glibc bug. Please reopen in that case.
Confirmed; it works!
How about SGI's XFS patches? Are they, or will they be included any time soon. (Perhaps I should add this as a separate issue?)
The same bug apparently reappears in kernel update 2.4.9-12, RedHat 7.1. Can anybody confirm this?
I've encountered similar problems with kernel-2.4.9-13 under Red Hat 7.2, but the symptoms aren't quite the same; with the original 7.1 kernel, one entry was missing in readdir() output on some, possibly all, directories, while on 2.4.9-13, *nothing* is returned for large directores. Refer to Bug 55864 for more info.