Bug 516298

Description of problem:
I have seen this problem with both F11 and rawhide (F12 Alpha).  My system has
two SATA drives, each of which is divided into a 12 or so partitions.  I have
created 11 md RAID-1 devices on top of these partitions:

  md0  (sda2,sdb2)    ext3 (/boot64)
  md1  (sda3,sdb3)    ext3 (/boot32)

  md2  (sda5,sdb5)    LVM PV
   .        .           .
   .        .           .
   .        .           .
  md11 (sda14,sdb14)  LVM PV

All of the PVs are part of a single volume group that holds LVs for various
root filesystems, KVM guest virtual disks, etc.

This worked just fine up through F10.  As of F11, however, this configuration
causes anaconda to crash with a message about "file descriptor out of range
in select()".  I cannot capture the backtrace, because that causes anaconda
to go back into its storage discovery code.

I suspect, although I am by no means sure, that anaconda may be making the
same assumption that mkinitrd did -- that sysfs will list all of the PVs
required by an LV's VG.  (See bug #506189.)

Version-Release number of selected component (if applicable):
anaconda from F12 Alpha DVD (x86_64)

How reproducible:
100%

Steps to Reproduce:
1. Set up system as described above.
2. Attempt to install Fedora 11 or rawhide (Fedora 12 Alpha).
  
Actual results:
Traceback

Expected results:
Successful install -- at least with Fedora 11.  ;-)

Additional info:
I'm more that willing to debug this further, but I don't know how to do so,
since there's no way for me to capture the backtrace.